University of Rhode Island DigitalCommons@URI Institute for and Informatics Faculty Institute for Immunology and Informatics (iCubed) Publications

2013 T-cell dependent immunogenicity of therapeutics: Preclinical assessment and mitigation Vibha Jawa

Leslie P. Cousens

See next page for additional authors

Creative Commons License

This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 3.0 License.

Follow this and additional works at: https://digitalcommons.uri.edu/immunology_facpubs

Terms of Use All rights reserved under copyright.

Citation/Publisher Attribution Jawa, V., Cousens, L.P., Awwad, M., Wakshull, E., Kropshofer, H., & De Groot, A.S. (2013). T-cell dependent immunogenicity of protein therapeutics: Preclinical assessment and mitigation. Clinical Immunology, 149(3.B), 534-555. Available at: http://dx.doi.org/10.1016/j.clim.2013.09.006

This Article is brought to you for free and open access by the Institute for Immunology and Informatics (iCubed) at DigitalCommons@URI. It has been accepted for inclusion in Institute for Immunology and Informatics Faculty Publications by an authorized administrator of DigitalCommons@URI. For more information, please contact [email protected]. Authors Vibha Jawa, Leslie P. Cousens, Michel Awwad, Eric Wakshull, Harald Kropshofer, and Anne S. De Groot

This article is available at DigitalCommons@URI: https://digitalcommons.uri.edu/immunology_facpubs/3 Clinical Immunology (2013) 149, 534–555

available at www.sciencedirect.com

Clinical Immunology

www.elsevier.com/locate/yclim

REVIEW T-cell dependent immunogenicity of protein therapeutics: Preclinical assessment and mitigation☆ Vibha Jawa a, Leslie P. Cousens b,1, Michel Awwad c,2, Eric Wakshull d, Harald Kropshofer e, Anne S. De Groot b,f,⁎ a Amgen, USA b EpiVax, USA c Pfizer, USA d Genentech, USA e Roche, Switzerland f University of Rhode Island, USA

Received 8 July 2013; accepted with revision 14 September 2013 Available online 25 September 2013

KEYWORDS Abstract Protein therapeutics hold a prominent and rapidly expanding place among medicinal Quality-by-Design; products. Purified blood products, recombinant , growth factors, enzyme replacement ; factors, monoclonal , fusion , and chimeric fusion proteins are all examples of Immunogenicity; therapeutic proteins that have been developed in the past few decades and approved for use in Cell-mediated the treatment of human disease. Despite early belief that the fully human nature of these proteins would represent a significant advantage, adverse effects associated with immune responses to some biologic therapies have become a topic of some concern. As a result, drug developers are devising strategies to assess immune responses to protein therapeutics during both the preclinical and the clinical phases of development. While there are many factors that contribute to protein immunogenicity, T cell- (thymus-) dependent (Td) responses appear to play a critical role in the development of responses to biologic therapeutics. A range of methodologies to predict and measure Td immune responses to protein drugs has been developed. This review will focus on the Td contribution to immunogenicity, summarizing

Abbreviations: Td, T-cell dependent, thymus dependent; T, thymus; ADA, anti-drug antibodies; Ti, T-cell independent; APC, -presenting cells; HLA, ; MHC, major histocompatibility complex; TCR, T cell receptor; Treg, regulatory T cells; FVIII, factor VIII; nTregs, natural regulatory T cells; aTreg, adaptive regulatory T cells; iTreg, induced regulatory T cells; IEDB, Immune

Epitope Database Analysis Resource; IC50, 50% inhibitory concentration; ELISpot, enzyme-linked immunosorbent spot-forming; ELISA, enzyme-linked immunosorbent assay; CFSE, carboxyfluorescein succinimidyl ester; PBMC, peripheral blood mononuclear cells; ALN, artificial lymph node; ORG, unmodified original ; FPX, recombinant Fc fusion protein; SFC, spot-forming cells. ☆ This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial-No Derivative Works License, which permits non-commercial use, distribution, and reproduction in any medium, provided the original author and source are credited. ⁎ Corresponding author at: 146 Clifford Street, Providence, RI USA. Fax: +1 401 272 7562. E-mail address: [email protected] (A.S. De Groot). 1 Denotes equal first-authorship. 2 Present affiliation: Merrimack Pharmaceuticals, USA.

1521-6616/$ - see front matter © 2013 The Authors. Published by Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.clim.2013.09.006 Td immunogenicity of biologics 535

current approaches for the prediction and measurement of T cell-dependent immune responses to protein biologics, discussing the advantages and limitations of these technologies, and suggesting a practical approach for assessing and mitigating Td immunogenicity. © 2013 The Authors. Published by Elsevier Inc. All rights reserved.

Contents

1. Introduction ...... 535 1.1. The immunogenicity of protein therapeutics ...... 535 2. The central role of T cells in immunogenicity ...... 537 2.1. The T cell contribution to antibody responses ...... 537 2.2. T cell stability and immunogenicity ...... 537 2.3. Td immunity to therapeutic proteins ...... 538 2.4. The role of central and to biologics ...... 538 3. Methods for predicting Td immune responses ...... 538 3.1. In silico T cell epitope-screening methods ...... 538 3.2. Strengths and limitations of in silico analysis...... 539 3.2.1. Antigen processing ...... 539 3.2.2. MHC affinity ...... 539 3.2.3. T cell phenotype ...... 540 3.2.4. Individual versus population-level predictions ...... 540 3.2.5. Post-translational factors ...... 540 3.3. HLA binding assays ...... 540 3.3.1. Competition binding assays ...... 540 3.3.2. Direct binding assays...... 541 3.3.3. Real-time kinetic measurements ...... 541 3.4. Strengths and limitations of HLA binding assays ...... 541 3.5. In vitro T cell assay methods for Td immunogenicity analysis...... 541 3.5.1. Measurement of T cell responses...... 541 3.5.2. T cell proliferation...... 542 3.5.3. Tetramers...... 542 3.5.4. Naïve T cell in vitro assays ...... 542 3.5.5. T cell assays using whole ...... 542 3.5.6. T cell re-stimulation assays using “exposed” donors ...... 542 3.5.7. Reconstitution of T cell immune responses in vitro ...... 542 3.6. Strengths and limitations of in vitro T cell assays for Td immunogenicity analysis ...... 543 3.7. Mouse models of in vivo Td immunogenicity of human therapeutic proteins ...... 543 3.7.1. HLA transgenic mice ...... 543 3.7.2. Humanized mouse models...... 544 3.8. Strengths and limitations of mouse models of in vivo Td immunogenicity ...... 544 4. Applied Td immunogenicity prediction ...... 545 4.1. In silico prediction supported by subsequent clinical data ...... 545 4.2. Clinical link between MHC class II haplotype and IFN-β immunogenicity ...... 545 5. Mitigation of T cell-dependent immunogenicity ...... 546 5.1. Deimmunization ...... 547 5.2. Tolerization ...... 547 6. Considerations: assessing immunogenicity of therapeutic proteins ...... 547 7. Conclusion ...... 548 Conflict of interest statement ...... 549 References...... 549

1. Introduction agents have entered the marketplace and have generated an estimated $99 billion in sales worldwide [1–4]. Therapeutic 1.1. The immunogenicity of protein therapeutics biologics offer the advantages of increased specificity and reduced toxicity compared to small molecules. However, when administered to patients, these protein-based drugs have the Since the approval of the first recombinant biological thera- potential to elicit immune responses that may directly impact peutic, insulin, in October 1982, more than 165 biotherapeutic drug safety, efficacy, and potency. For example, anti-drug 536 V. Jawa et al. antibodies (ADA) that develop in response to a therapeutic into patient-, disease-, or product-related categories. The protein may alter the drug's pharmacokinetic profile and patient- and disease-related categories describe factors abrogate its pharmacodynamic effect (neutralizing activity) that may pre-dispose a particular individual to an immune [5–8]. response. In contrast, product-related factors, i.e. factors Immune responses to proteins are characterized by the intrinsic to the final drug product itself that contribute generation of antibodies (humoral immune response) that to immunogenicity, may include modifications in the could be T cell dependent or independent. T-independent glycosylation profile [18–20], biophysical and biochemical antibody responses may be generated when B cells recognize attributes [21–24], or factors introduced during formulation a repeated pattern (motif) in the therapeutic protein and [14,20,25,26]. Table 1 summarizes some of these patient-, respond by transiently producing low-affinity, predominate- disease-, and product-related factors that have the poten- ly IgM antibodies [9]. Antibodies that are generated in tial to influence immune responses to a biologic therapeutic. conjunction with T cell help are referred to as T cell- Many of the factors described in Table 1 that might dependent or thymus-dependent (Td) antibodies. This predispose a therapeutic protein to be immunogenic have process, described in the next section, involves a complex also been identified as “critical quality attributes” in the interplay among antigen presenting cells, T cells, secreted FDA-sponsored Quality-by-Design initiative [27] focused on cytokines, and B cells, emphasizing the importance of manufacturing “process development”. For protein-based genetic factors such as HLA haplotype expression and T drugs, the primary amino acid sequence itself can be a cell/ repertoire in immune responses to administered strong determinant of immunogenic potential. Unfortunate- proteins. Thus, measurement of ADA IgG responses usually ly, primary sequence is difficult, if not impossible, to modify indicates that T cells are involved in the immune response to in the manufacturing process. the protein. Moreover, a number of clinical studies now Although many protein drugs are associated with ADA, as suggest that high levels of T cell-driven IgG ADA have the mentioned above, only a few have been associated with potential to cross-react with the endogenous counterpart, severe adverse events such as profound anemia or thrombo- an adverse effect that can have serious consequences cytopenia when the antibodies cross-reacted with the drug's [10–12]. endogenous counterparts [10,11,28]. These types of serious T cell responses contribute to the generation of ADA. outcomes resulting from cross-reactive ADA have inspired Proteins with therapeutic potential are produced in cell lines the development of a multitude of in vitro methods for that are derived from a variety of sources, including measuring the presence of ADA, which have been carefully mammals (both human and non-human), insects, bacteria, reviewed in several white papers and regulatory guidance plants, yeast, and viruses. Small differences in the protein documents [17,22,29–33]. In addition, methods for identi- sequence and/or post-translational modifications (e.g., fying drivers of Td immune responses have also expanded glycosylation, oxidation, deamidation, acylation, and alkyl- and have been described in a number of publications ation) can contribute to the immunogenic potential of the [34–42]. therapeutic protein. Furthermore, the manufacturing pro- The goal of this paper is to review recent developments in cess may involve multiple steps (e.g., regulation of gene the understanding and identification of Td immune response expression, purification, concentration, formulation, and long-term stabilization) to produce recombinant proteins of sufficient quantity and quality to meet clinical release Table 1 Factors that can influence immune responses to a criteria. At each of these steps, there is potential for the therapeutic protein. introduction of biochemical or biophysical modifications into FACTORS IMPLICATIONS the molecules that may influence the immunogenicity Intrinsic immunogenic dominant epitopes Higher risk of immunogenicity based on profile of the biologic product. While these aspects of reactivity to HLA (DR, DP, DQ) alleles protein production may not impact function of the thera- Sequence differences between therapeutic Cross–reactive neutralizing antibodies protein and endogenous protein to endogenous proteins peutic protein, modifications inherent in the manufacturing Differences in glycosylation patterns. Structural Breaking of tolerance due to process can have a major impact on host immune responses. alterations induced by storage condition, – Aggregation production/purification and formulation – Oxidation For example, consider the case of Erbitux, a protein with – Deamidation and degradation non-human type glycosylations that caused anaphylaxis in – Conformational changes Route of delivery selected patients who had pre-existing antibodies targeting – Route/frequency of administration – Differences in following intravenous and subcutaneous these carbohydrate structures [13,14]. Host cell proteins delivery derived from recombinant protein-producing cell lines may – Syringe environment and associated – Presence of leachates from syringe (i.e. reconstitution tungsten or silicone) can facilitate immune co-purify with a therapeutic product to become part of the responses – Non–physiologic concentrations – Exposure to high dose proteins in final formulation [15,16]. These impurities, even in small formulations can lead to an immune quantities, have the potential to stimulate an unwanted response Patient–related factors immune response. In at least one recent case, they induced • Immune status – Immune status can enhance or reduce – Immunosuppression due to the likelihood of an immune response anti-host cell protein antibodies and contributed to the supportive therapies – Presence of HLA alleles that have high suspension of a clinical trial [15,16]. – Autoimmune–diseased state potential to bind to immune dominant – Genetic features like HLA epitopes So as to provide biologics developers with a structured haplotypes • Intensity of treatment at the first dose approach to measuring immunogenicity risk, the European • Age Medicines Agency (EMA) has published a “Guideline on Host Cell Proteins (HCP) Immunogenicity Assessment of Biotechnology-Derived Ther- The homology of HCP with human proteins Recombinant protein therapeutics produced raises concern over the inherent risk of apeutic Proteins” [17], in which factors influencing the in mammalian cell lines may contain anti–self immune responses. immunogenicity of therapeutic proteins were classified process–related contaminants. Td immunogenicity of biologics 537 drivers, and to organize these into a framework to approach sufficiently high stability with MHC will facilitate presenta- the prediction and measurement of Td immune responses as tion of the peptide/MHC complex on the APC cell surface3 one component of a risk assessment strategy. Case studies [50]. Because human populations express a number of will be presented to illustrate Td immunogenicity and the different HLA class II alleles, the interaction between practical application of these methods. An evidence-based antigenic epitope and HLA may exhibit a full spectrum of roadmap is proposed for identifying Td responses in protein binding stabilities. HLA polymorphism and its impact on the therapeutics and developing criteria for assessing immuno- binding of specific peptides (HLA restriction) are primary genicity by (i) sequence-driven assessment using in silico mechanisms by which patient genetics contributes to algorithms, (ii) in vitro assays, and (iii) in vivo models. immune responses to particular protein therapeutics. Exam- Lastly, we touch on the emergence of methods for mitigating ples of HLA-associated immune responses to therapeutic Td immunogenicity, such as de-immunization and tolerance proteins are provided below. induction. CD4+ T cell recognition of a specific antigen-MHC class II complex, combined with co-stimulatory signals delivered by the APC, culminates in robust T cell activation that in turn 2. The central role of T cells in immunogenicity facilitates a mature antibody response. In the absence of activated T helper cells, naïve B cells do not fully mature, 2.1. The T cell contribution to antibody responses and activated antigen-specific B cells are rendered anergic or undergo apoptosis. Therefore, T cell recognition of peptide epitopes derived from a protein therapeutic is the Immune responses to therapeutic proteins can be broadly key determinant of Td antibody formation. divided into two general categories. The first involves activation of the classic adaptive by what are considered to be “foreign” proteins, like the response 2.2. T cell epitope stability and immunogenicity elicited against pathogens, vaccines, or allotypic antigens. Replacement proteins would be viewed as foreign to the Immunogenicity prediction is based on well-defined interac- immune system of an individual who is lacking, in whole or in tions between the amino acids comprising a protein part, the endogenous counterpart. The second involves a sequence and an individual's HLA molecules. Setting aside breach of B and/or T cell tolerance, like the response the influence of T cell receptor (TCR) affinity, the greater elicited to autologous self-proteins in certain autoimmune the stability of a given peptide within the binding groove of a diseases. The mechanisms for the breach of immune particular HLA class II molecule, the greater the likelihood tolerance are not as well defined as the mechanisms for that this peptide will elicit a T cell response [50]. Thus, the immune induction to microbial infections, but may include identification of peptide sequences in a biologic drug that epitope mimicry, cross-reactivity of T cells, presence of bind to HLA is highly relevant for achieving the goal of trace levels of innate immune activators such as toll-like screening for immunogenic potential. receptor agonists [25,43,44], and/or multivalency as might The presence of HLA-binding sequences in a protein be presented by aggregated proteins [45]. therapeutic is part of the larger biological process of antigen For the classic immune pathway, production of anti- presentation, T cell receptor (TCR) recognition, and T cell therapeutic protein responses is the culmination of a series activation, and thus is not an absolute guarantee that T cell of events that eventually leads to B cell activation and responses will occur. As elegantly described by others, antibody secretion. This antibody production can be thymus peptide processing and abundance, intracellular editing and independent [T-cell independent, (Ti)] or Td in origin MHC loading by the DM protein, peptide–MHC stability, TCR [46,47]. B cells are activated in a Ti manner when particular specificity and abundance, phenotype of the responding T structural patterns, such as polymeric repeats or carbohy- cell, and presence or absence of secondary signals all drate molecules, directly induce activation of B cells. Ti contribute to initiating and shaping the quality and quantity activation of B cells can be distinguished from Td activation, of a T cell response [18,20,25,26,50]. While the detection of as the antibodies resulting from Ti activation are limited in a T cell response to a putative epitope is an important both and affinity and if memory B cells are generated predictor of immunogenic potential, it has been much more they are not long-lived [48,49]. In contrast, Td activation of difficult to reliably model the intricate yet critical steps that B cells is characterized by class switching (IgM to IgG) and occur between the HLA binding event and the activation of T development of memory B cells that produce higher-affinity, cell response. Furthermore, the important contribution of more robust, and longer-lived antibody responses. The regulatory T cells (Treg), which may proliferate and produce development of IgG-class antibodies following administra- inhibitory signals in response to particular T cell epitopes, tion of a biological protein generally indicates that the has gained appreciation in recent years. Methods for therapeutic is driving a Td immune response. detecting and discriminating between regulatory and helper Td responses, by definition, are contingent upon T cell T cell responses (Treg and T helper, respectively) are recognition of therapeutic protein-derived epitopes through discussed in greater detail below. Quite a few clinical trials the basic processes of protein antigen processing and of protein therapeutics have now linked the presence of T presentation. The therapeutic protein is taken up by helper responses to immunogenicity, confirming the direct antigen-presenting cells (APC) and proteolytically processed into small peptides. Certain of these peptides will bind to human leukocyte antigen (HLA)/major histocompatibility 3 The terms HLA (human) and MHC (encompasses both human and complex (MHC) class II molecules, where interactions of mouse) are used interchangeably in this manuscript. 538 V. Jawa et al. relationship between T cell epitopes and the immunogenic- 3. Methods for predicting Td immune responses ity of biologics [6,39,51–54]. Detailed knowledge of the regions of localized immunoge- nicity within the protein sequence may facilitate immuno- 2.3. Td immunity to therapeutic proteins genicity mitigation efforts. Here we begin with a review of immunogenicity prediction methods, followed by a brief A classic T cell and Td antibody response to a therapeutic description of strategies that may mitigate Td immunoge- protein antigen can occur after one or two administrations nicity including tolerization, deimmunization, and the of a particular protein therapeutic, persist for prolonged coadministration of the drug with immunosuppressive periods of time, and ultimately inhibit the efficacy of the therapies such as methotrexate, prednisone, cyclophospha- therapeutic product. A classic Td immune response to a mide, or Rituxan [65–68]. protein therapeutic is illustrated by the induction of ADA Predictive immunogenicity screening often involves more “inhibitors” in response to blood replacement factor VIII than one approach, as each method has strengths and (FVIII) in hemophilia patients. Although the incidence and weaknesses. A first step in the process may be to screen a intensity of the immune response to FVIII can vary depending protein for the presence of T cell epitopes by sequence on the extent to which endogenous FVIII is expressed in the analysis in silico. This step can be followed by in vitro studies individual patient, immune responses to FVIII are driven by T using a variety of methodologies, including HLA binding cell epitopes [55]. The differential between effector and assays, naïve blood assays, or humanized mouse models (see immune responses to FVIII epitopes, on the below for descriptions of these assays). individual level, may be a factor that determines whether that individual develops anti-FVIII ADA [56]. 3.1. In silico T cell epitope-screening methods

The core residues of the T cell epitope sequence that mainly 2.4. The role of central and peripheral tolerance to define the affinity and stability of binding to HLA pockets are biologics limited in length to 9–10 amino acids; thus prediction of T cell epitopes based on the amino acid sequence of a peptide is The absence of an immune response to autologous proteins is computationally feasible when sufficient information on a set of attributed to central (thymus-derived) to peptides that are known to bind to a particular MHC is available. proteins present in the respective individual's proteome. Databases such as the Immune Epitope Database Analysis According to the theory of , T cells that Resource (IEDB; www.tools.immunoepitope.org) [69] provide respond to epitopes derived from autologous proteins the raw material for developing T cell epitope prediction tools. expressed in the thymus undergo deletion or are rendered Presentation of epitopes by both MHC I and MHC II anergic during thymic development [57]. Nonetheless, contributes to the initiation of an immune response. MHC II tolerance to autologous proteins is now known to be is more relevant to anti-drug antibody responses, as MHC incomplete, since autoreactive T cells to self-antigens have Class II-restricted T helper cells are responsible for driving been uncovered in peripheral circulation in the context of humoralimmunity.HLADR,DQ,andDParethethreeloci and are also present in the circulation of of peptide-carrying HLA class II molecules. In vitro and healthy individuals. Natural regulatory T cells (nTregs) are T in vivo observations indicate that HLA DR binding peptides cells generated in the thymus that circulate in the are generally 12 amino acids in length, as the flanking periphery. Upon antigen-specific activation through their sequences serve to stabilize the peptide (by hydrogen TCR, CD4+CD25+FoxP3+ nTregs are able to suppress bystand- bonds) in the HLA binding groove [70]. Systematic assess- er effector T cell responses to unrelated antigens by ments of MHC class II peptide binding domains of proteins contact-dependent and -independent mechanisms [58,59]. have been developed using machine learning methods, nTregs may hold immune responses to some autologous software algorithms, and data transformations [71–73]. proteins (such as erythropoietin) in check. For a brief overview of T cell tools, see Adaptive Tregs, developed in reaction to foreign antigens Table 2 [69,74–80]. A common denominator among these to which central tolerance may not exist, are an additional tools is the ability to quickly screen large datasets, source of control over immune responses [60]. Sustained including whole genomes or proteomes, for putative T cell tolerance (to exogenous proteins) may require the existence epitopes. Several common HLA-DR types share HLA binding of these ‘adaptive’ or ‘induced’ Treg cells (aTreg, also pocket repertoires [81], meaning their ability to bind known as iTreg, [58]) with the same antigen specificity as peptides is more promiscuous than for class I alleles. the self-reactive effector T cells [58,61–63]. Administration Moreover, analysis focused on as few as eight representa- of antigen in the absence of an innate-immune stimulator tives of the 875 known HLA-DR alleles can “cover” the (danger signal) can lead to tolerance; this approach has been genetic backgrounds of most humans worldwide [82]. used for the induction of tolerance to . A strong link Additionally, MHC class II-binding T cell epitopes have been connecting HLA, presentation of T cell epitopes (both observed to occur in clusters of up to 25 amino acids in length regulatory and effector) in the context of HLA, and the [72,83]. Thus identifying MHC class II-binding T cell epitope maintenance of peripheral tolerance has been described clusters can be a strong indication for T cell responses because [64]. The implications of immune control by nTregs and the they represent regions of the protein in which sequences that induction of tolerance via aTregs will be discussed in greater have high affinity across multiple HLA alleles and multiple detail below. frames are located [79]. Td immunogenicity of biologics 539

It is notable that there are few examples in the literature typically by at least 20-fold [98], and provides an opportu- of an association between HLA-DP or -DQ and immunogenic- nity in early development to modify a protein to decrease its ity. This may be due to the fact that binding motifs of immunogenic potential. HLA-DP and HLA-DQ alleles are less well defined than HLA-DR alleles, and/or that HLA-DP and HLA-DQ may 3.2. Strengths and limitations of in silico analysis contribute very little to drug-related immunogenicity. A recent retrospective analysis of previously published immu- 3.2.1. Antigen processing nogenicity prediction data [72] found no association be- In vivo, immunologically relevant epitopes are produced by tween the presence of DP or DQ motifs (as predicted using the APC, which processes the protein into discrete peptide online tools accessible through the IEDB) and immunogenic- fragments, assembles the peptides in complexes with MHC ity of 23 monoclonal antibodies that have been evaluated in molecules, and displays these complexes on its cell surface human clinical trials [199]. [99,100]. The non-classical MHC class II molecule HLA-DM The computational approach to MHC class II acts as a T cell epitope editor, reducing the number of epitope prediction and its application to Td immunogenicity epitopes that are eventually presented at the APC surface assessment has been described in previous publications through kinetic proofreading [101]. Methods that will [84–90] and is now well accepted in vaccine discovery predict products of antigen processing and contribute to efforts [85,91–94] and for the identification of key epitopes the accuracy of Td immunogenicity assessments are under triggering autoimmunity [95]. In the context of biologics, development [50,102,103]. One means to qualitative- several studies of therapeutic proteins mentioned above ly identify the peptide sequence of epitopes that are [6,51] have demonstrated the prospective predictive accu- presented following HLA-binding and DM-editing is to racy of these algorithms by showing a positive correlation directly elute naturally processed HLA-associated peptides, between the presence of an MHC binding sequence and followed by peptide sequencing [103–105]. observed Td immunogenicity in the clinic (demonstrated for a recombinant fusion protein, FPX, in Table 3). A specific 3.2.2. MHC affinity example is provided in detail in the case studies section of A prerequisite for T cell epitope immunogenicity is the this article. Some groups have also developed tools that are stable binding of a linear epitope in the groove of the MHC expressly designed to rank therapeutic proteins by T cell molecule; the higher the kinetic stability (slow peptide epitope content [72] and to define means of modifying off-rate), the longer it will reside within the MHC cleft, and epitope content in silico [35,96,97]. Application of in silico the greater the likelihood that an epitope will be recognized tools reduces downstream in vitro testing dramatically, by a T cell [106]. Improved prediction of peptide–MHC

Table 2 Summary of T cell epitope mapping tools, in alphabetical order.

NAME DEVELOPERS/INSTITUTION TYPE WEBSITE

Epibase I. Lasters and P. Stas Commercial www.lonza.com Algonomics NV/Lonza, Inc.

EpiMatrix A.S. De Groot and W.D. Martin Collaborative/Commercial www.epivax.com EpiVax, Inc.

IEDB Vita R, Zarebski L, Greenbaum JA, Emami H, Hoof I, Mixed collection of www.iedb.com Salimi N, Damle R, Sette A, Peters B. The immune epitope tools of assorted derivation database 2.0. Nucleic Acids Res. 2010 Jan;38:D854-62.

MHC2PRED G.P.S. Raghava Public www.imtech.res.in/raghava/mhc2pred/ Bioinformatics Center, Institute of Microbial Technology, Chandigarh, India

MHCPRED D.R. Flower Public www.ddg-pharmfac.net/mhcpred/MHCPred/ The Jenner Institute

PROPRED/ G.P.S. Raghava and H. Singh Public www.imtech.res.in/raghava/propred/ TEPITOPE Bioinformatics Center, Institute of Microbial Technology, Chandigarh, India

RANKPEP P.A. Reche Public http://bio.dfci.harvard.edu/RANKPEP/ Harvard Medical School

SVRMHC P. Donnes, A. Elofsson Public http://svrmhc.biolead.org/ Division for Simulation of Biological Systems, University of Tubingen, Germany

SYFPEITHI H.G. Rammensee Public http://www.syfpeithi.de Department of Immunology, Tubingen, Germany

SMM-Align/ M. Nielsen, C.Lundegaard, and O. Lund Public www.cbs.dtu.dk/services/NetMHCII-2.2/ NetMHCII-2.2 Center for Biological Sequence Analysis, Department of Systems Biology, Technical University of Denmark

TCED/iTope M. Baker and F. Carr Commercial www.antitope.co.uk/ Antitope, Ltd. 540 V. Jawa et al.

Table 3 Association of algorithm-predicted immunogenicity peptide sequences do not induce T cell activation [109]. to clinical immunogenicity rates. EpiMatrix-generated scores Thus, Td immunogenicity on the population level, as associated with each FPX protein and their respective rates of predicted by tools that measure the potential for immuno- antibody incidence (binding and neutralizing) are shown. An genicity across HLA alleles, can be quite different from assessment of Tregitope content in each molecule was also immunogenicity at the individual level. In silico tools can be performed and scores were adjusted accordingly. FPX 1, for used to predict the former in the pre-clinical phase of example, had a high rate of clinical immunogenicity and was development; the same tools may be useful to analyze associated with elevated T-cell epitope content and low subject-by-subject immune responses to a protein thera- Tregitope content, as reflected by its high Z score. FPX 2, 3, peutic in the clinical phase of development. and 4 were associated with a low EpiMatrix score, and Tregitope adjustment further reduced the predicted poten- 3.2.5. Post-translational factors tial for binding. Predictably, FPX 2, 3, and 4 exhibited only Additional determinants of Td immunogenicity that are not minor clinical immunogenicity. considered in the in silico prediction process may be

Protein Therapeutics FPX 1 FPX 2 FPX 3 FPX 4 revealed in vitro. For example, effects of non-sequence- derived post-translational modifications including those of a EpiMatrix Score 21.97 1.76 —0.76 1.63 biophysical or biochemical nature such as deimination (or Tregitope — adjusted citrullination), deamidation, oxidation, dimerization, and 21.97 1.62 —1.76 —111.25 EpiMatrix score protein folding-induced conformational changes that are Binding Antibodies 37% 7.80% 5.60% 4.50% all widely accepted as determinants of immunogenicity [21,110]. Neutralizing Antibodies 40% 0.50% Not Analyzed 0% Despite these caveats, T cell epitope prediction has advanced greatly over the last decade. In silico tools provide the advantages of high throughput, low cost, and the ability to reduce the search space for further in vitro testing. In vivo stability might improve the ability of in silico methods to immunogenicity, however, is also influenced by a number of predict activation of CD4+ T cells in vivo [106]. While in silico factors that cannot yet be accurately predicted in silico tools have dramatically improved over time, and consider- [71,80]. The results of in silico analyses can then be carried ation of epitope “off-rate” or peptide “residence time” on to the next step of pre-clinical analysis, such as HLA binding, the MHC molecule is probably within reach of future antigen processing and presentation assays, and T cell generations of algorithms, in contrast, TCR affinity is assays, in which other factors that contribute to generating unlikely to be amenable to in silico prediction, as the TCR an immune response can be examined. is pleiomorphic and exhibits three-dimensional complexity akin to B cell epitopes. 3.3. HLA binding assays

3.2.3. T cell phenotype Having defined the regions of interest in silico, one means of Identification of T cell epitopes in a protein may indicate validating these predictions is to perform HLA/MHC binding that the protein could induce an immune response, though assays. A number of different HLA class II binding assays can the nature of the response remains heavily dependent on the be used to measure the affinity of predicted epitope phenotype of the T cell that recognizes this epitope in the sequences to HLA alleles in vitro; several are described in context of MHC class II and on the local milieu (inflammatory this section. or regulatory). The prediction of the type of T cell (T helper, T regulatory, etc.) responding to a potential T cell epitope 3.3.1. Competition binding assays remains beyond the capabilities of the in silico approaches One method that is employed is a competition-based HLA described here: phenotyping requires in vitro characteriza- binding assay, described previously [111–113] that can be tion, except in cases where known regulatory T cell epitopes adapted for high throughput [114]. The method is described have been previously described. in detail in other publications, but briefly, experimental peptides are tested for their ability to compete against a 3.2.4. Individual versus population-level predictions labeled high-affinity peptide for binding to purified class II Caution is advised when applying T cell epitope prediction molecules. The reaction is carried out at several peptide tools that have been developed for the purpose of predicting concentrations; a non-linear regression analysis of the immune responses in large populations to an individual resulting dose–response curve is then used to calculate the subject. As HLA binding is intimately associated with T cell IC50 (concentration of test peptide required to compete off response, individual responses to protein therapeutics will 50% of the target peptide). Binding assays can performed for vary depending on multiple factors, such as an individual's a broad representation of class II alleles [82]. One drawback HLA haplotype [53,89] and immune status. Individual T cell of this method is that measurements of peptide affinity are repertoire variation further complicates the interpretation always relative to this standard reference peptide, as of individual subject responses; current thinking is that a T the appropriate reference peptide may differ from HLA cell repertoire is defined by previous exposures to related allele to allele. A value that describes the 50% inhibitory epitopes, vaccination, and the internal (gut) microbiome concentration (IC50) can be derived by fine-tuning the [107,108]. Such individual specificity, in addition to antigen protocol for each allele through repeated experiments, and processing (see above), may explain why some MHC binding performing the assays at multiple concentrations. IC50 values Td immunogenicity of biologics 541 can be converted to method-independent Ki values via the While epitopes can be predicted (in silico) and validated Cheng–Prusoff equation for comparison of peptide affinities in binding assays (in vitro), the final impact of the T cell across platforms [115]. epitope is through its activation of a T cell. Activation is related to TCR specificity, avidity and T cell phenotype (none of which are measured in HLA binding assays), which 3.3.2. Direct binding assays makes straightforward interpretation of immunogenicity It is also possible to measure peptide affinity directly, rather based upon HLA binding assays impossible. Thus, HLA binding than observing its ability to displace a known ligand. In such assays may improve the accuracy of immunogenicity predic- experiments, a variety of methods may be used to separate tions when applied in a step-wise process after in silico peptide bound to HLA molecules from free peptide in epitope prediction and before the employment of a solution, thus allowing for assessment of the peptide's biological assay such as enzyme-linked immunosorbent affinity. Spin-column filtration and gel electrophoresis are spot-forming (ELISpot) assays or HLA transgenic mouse ways to accomplish this distinction by molecular weight, studies, but additional assays that assess T cell responses since peptide–HLA complexes will weigh more than unbound may need to be performed [120]. peptides. These methods, however, are tedious and can suffer from complex dissociation during the separation process. Other novel methods have been proposed, such as 3.5. In vitro T cell assay methods for Td proximity-based detection utilizing a Luminescent Oxygen immunogenicity analysis Channeling Immunoassay [116]. All of these techniques allow more direct binding affinity determination than competition studies, but have methodological limitations and require For many years, in vitro assays based on HLA haplotype have costly equipment. been used in transplantation research to assess the risk of engrafted T cells responding to host tissue (graft-versus-host disease). Adaptation of these assays to protein therapeutics 3.3.3. Real-time kinetic measurements may improve the pre-clinical assessment of the potential for Apart from determining the quantity of displaced reference Td immunogenicity. The presence of T cells that actively peptide (competition assay) or bound vs. free test peptide suppress immune responses to autologous proteins is a (direct assay), the rate at which peptides interact with significant confounding factor in their development and HLA molecules is a dimension of epitope strength that can evaluation. In addition, these assays are dependent on the be measured in vitro. Protocols based on fluorescence selection of a culture milieu that accurately accounts for the polarization are suitable for kinetic studies due to the in vivo conditions of human immune stimulation [40,54,121]. absence of interference from ELISA wash steps, allowing In this respect, bulk cultures of PBMC, either with or without multiple readings to be taken throughout the binding co-stimulatory signals (anti-CD28 antibody, IL-2, IL-7, etc.), reaction [117].Additionally,surfaceplasmonresonance have been utilized to assess immunogenicity of therapeutic methods have also been employed to measure peptide proteins. binding to MHC II [118]. A number of biological outcomes for T cell activation can be measured in these in vitro assays, including cytokine secretion (IFN-γ, IL-2, IL-4, etc.), regulation of cell surface 3.4. Strengths and limitations of HLA binding assays markers of activation, signal transduction events, and proliferation [122,123]. Supported by such evidence, in In vitro HLA binding assays are relatively straightforward and silico-identified peptides that stimulate multifunctional T easier to perform than live-cell experiments. However, each cell responses in vitro can be considered bona fide T cell of the assay formats can be affected by peptide purity and epitopes with the potential to contribute to an ADA length as described here. Peptides that are synthesized for response. HLA binding assays must include carboxy- and amino- terminal flanks that stabilize the peptide in the MHC groove [119]. Selection of peptides based on artificially defined 3.5.1. Measurement of T cell cytokine responses overlapping sequences (e.g., 15mers overlapping by ten ELISAs and ELISpots are two related methods of measuring amino acids) can lead to the truncation of MHC binding cytokines secreted by T cells (i.e., IFN-γ, IL-2, IL-4, and motifs and elimination of critical flanking regions, which IL-10). The enzyme-linked immunosorbent assay (ELISA) can limits the accuracy of the binding assay. Furthermore, measure cytokine levels in culture supernatants generated artificially synthesized peptides may contain impurities, under conditions of T cell stimulation. Cytokine levels truncations, and errors in the sequence that can lower or measured in an ELISA can provide information about the alter the binding of the pure peptides; thus high-quality magnitude of the response (how much cytokine is secreted peptides (greater than 90% purity) are required to obtain into the supernatant) as well as the type and quality of the more accurate results in binding assays. Long peptides can response (which cytokines are or are not secreted into the fold or peptides can aggregate in solution, leading to supernatant). Multi-analyte, high throughput cytokine test- underperformance in binding assays. Solvents used in the ing can also be performed in cell-derived supernatants in experiment may interact negatively with certain amino bead-based assays [124–127]. Enzyme-linked immunosor- acids, causing oxidation or the formation of unwanted bent spot-forming (ELISpot) assays provide information disulfide bonds. Quality-controlled peptide manufacturing regarding the number of cytokine-producing cells (down to and storage, along with proper reagent selection, can 1 cell per million) within a cell population stimulated ex minimize the impact of these problems [113,114]. vivo; these assays are considered to be more sensitive and 542 V. Jawa et al. quantitative whereas ELISA and bead-based assays are 3.5.5. T cell assays using whole antigens slightly less sensitive and more qualitative. T cell assays using whole antigens also can be used to measure T Intracellular cytokine staining measured by flow cytom- cell responses to protein therapeutics in vitro. The recognition etry is another method for detecting cytokines and linking of these antigens requires the presence of an APC that is their expression to the phenotype of individual cells. capable of processing and presenting peptides derived from the These assays can be used to accurately measure T cell antigen. Advantages of whole PBMC assays include the ability polyfunctionality relative to the phenotypic classifications to set up several assays and/or assay conditions with a limited of CD4+ T cells based on cell surface markers. blood sample volume and the ease of assay performance, features which lend themselves to high-throughput assay 3.5.2. T cell proliferation development [40]. Human -derived dendritic cells T cell proliferation in response to stimulation by a peptide– can be manipulated in vitro to model antigen processing by MHC complex can be measured by (1) the incorporation professional APC in vivo (H. Kropshofer, unpublished data). of the radioactive nucleotide tritiated thymidine into the However, as applied to evaluating immunogenicity of biologics, DNA of dividing but not resting cells or (2) the dilution of optimization of variables such as the ratio of DCs to autologous a fluorescent dye, carboxyfluorescein succinimidyl ester T cells are important factors for ensuring that the in vitro (CFSE), that decreases in fluorescence intensity by half results are relevant to the clinical scenario. with each round of cell division and can be measured by flow cytometry. Thymidine incorporation assays are gradually 3.5.6. T cell re-stimulation assays using “exposed” donors being replaced by CFSE staining, which presents significant T cells re-stimulation assays are generally used to identify advantages in terms of ease-of-use. In addition to CFSE labeling, and measure a recall or memory response in PBMC derived cells can be co-stained for expression of other cell surface from subjects who have been exposed at a distant time point markers, transcription factors, and/or intracellular cytokines to a protein or a given biologic product. Whereas ‘exposed that distinguish between regulatory T cells and effector blood’ assays cannot be performed for novel therapeutics, T-helper cell phenotypes, including Th1, Th2, Th17, Th22, this type of assay can be used to evaluate the impact of etc. as reviewed in detail elsewhere [128–130]. pre-existing immune responses to a new version of an existing biologic in use such as a re-engineered FVIII. Epitope 3.5.3. Tetramers mapping of the recall response can be performed using Fluorescently labeled tetrameric complexes of MHC class II specific peptides from the whole protein; however, this molecules loaded with the peptide of interest (i.e., “tetra- approach may be over-predictive. Therefore, studies using mers”) can also be used to enumerate T cells recognizing a whole therapeutic protein should ideally be performed in particular epitope. However, relative to MHC I tetramers, MHC parallel with studies using sets of predicted peptides. II tetramer staining has proven problematic and so their utility Antigen-specific T cell responses can be assessed after has been restricted. The technical limitations experienced re-stimulation ex vivo by ELISA, ELISpot, proliferation, and specifically with MHC II tetramers may be due to weaker TCR– flow cytometric analysis of activation markers and intracel- peptide–MHC II interactions or the fact that CD4 does not lular cytokine expression. participate in the stability of tetramer binding as does CD8 [131]; either one or both of these features may contribute to 3.5.7. Reconstitution of T cell immune responses in vitro increased variability and poor quality of the experimental New methods for “reconstituting T cell immune responses” output. Efforts continue to improve MHC II tetramer–TCR in artificial media may permit improved in vitro assess- interactions towards the goal of increasing the utility of MHC II ments of the interactions between the professional APC and tetramers so that they may be more widely used for the the T cells. To this end, several artificial lymph node (ALN) identification of specific CD4 T cell responses, including those systems have been developed [135–137]. These methods against protein therapeutics. attempt to replicate, in three dimensional structures and in APC:PBMC ratios, the natural immune environment [138].In 3.5.4. Naïve T cell in vitro assays at least one approach, human blood-derived dendritic cells Naïve peripheral blood mononuclear cells (PBMC) have been are cultured in transwells partitioned by human vascular used to describe Td immunogenicity of therapeutic proteins endothelial cells. Addition of autologous CD4+ T cells to the [34,40,132,133]. Compared to a recall response, the precursor co-culture allows activated APC expressing chemokine frequency of antigen-specific cells in a naïve population is receptors to respond to inflammatory chemokines and quite low; it was postulated that the higher the precursor migrate through the transwells, mimicking the migration frequency reactive with a certain peptide or protein, the of APC from the periphery to the lymph nodes. The higher the potential of the respective peptide or protein to antigen-specific CD4+ T cell response is monitored by induce an immune response [134]. Multiple rounds of antigen CD154, IFN-γ, IL-2, IL-5, IL-17, and IL-21 expression. A stimulation, sometimes over several weeks, are required to good correlation between previously established immune expand sufficient numbers of antigen-specific T cells for responses in vivo and ALN immunogenicity has been reliable measurement. By conditioning naïve blood samples ex observed, at least for protein-based vaccines [139].Even vivo through prolonged and/or repeated exposure to experi- though the human ALN model is primarily being used for in mental antigen, immune responses can be expanded to the vitro evaluation of vaccine efficacy [140], the application of point where they can be measured. What is not known is how this technology for the prediction of therapeutic protein expansion affects the phenotype of the T cell response; either immunogenicity is feasible. regulatory or effector responders may eventually dominate in Alternatively, some groups have used small flow-through in vitro cultures. systems that induce PBMC to self-assemble into lymph Td immunogenicity of biologics 543 node-like structures [141]. Pre-activation of the T cells Similarly, stimulation can be performed with peptides or improves the readouts; however, the introduction of whole proteins alone, peptides in the context of tetramers, specialized (non-self) APC may generate false positives for or APC pulsed with whole proteins. New advances in some biologic proteins. In other cases, these ALN systems tetramer/multimer technology can enhance detection of have given results that correlate with predicted immune epitope-specific T cells and thus should allow more sensitive responses [135,142]. and standardized approaches to evaluate individual re- sponses to T cell epitopes identified in protein therapeutics 3.6. Strengths and limitations of in vitro T cell [145]. assays for Td immunogenicity analysis Finally, the number of individual blood (PBMC) donors that would normally be required to address the HLA diversity of a patient population is quite large (more than forty), and The benefit of in vitro T cell assays utilizing easily accessible the volume of blood required ranges from 15 mL to more human peripheral blood is that the assays may provide a than 50 mL. Maintenance of a large supply of blood samples preview of the immunogenicity of a therapeutic protein from pre-qualified donors that is sufficient to reduce without the risks typically associated with first-in-human assay-to-assay variation can be done, but is cost prohibitive use. Careful consideration of the composition of the cells in for most preclinical laboratories. culture, such as monocyte-derived , immature/ Much remains to be done to improve the accuracy of in vitro mature dendritic cells, and T and B cells, may improve the T cell assays in predicting clinical immunogenicity. Future ability of a protein to initiate and propagate an immune considerations for improving in vitro T cell assays include: response. Furthermore, these assays incorporate antigen better linkage between the spectrum of immune responses to a processing and presentation pathways for whole proteins as therapeutic protein and predictive power in clinical trials well as for discrete peptides. using statistically derived criteria, such as fold-increase or A limitation of ELISpot and ELISA assays is that PBMC contain stimulation-index; improved means to distinguish responders several cell types capable of secreting particular cytokines, from non-responders; evaluation of T cell responses from e.g., IFN-γ (NK cells, NKT cells, CD4+ or CD8+ T cells). Thus diseased states associated with or immune PBMC-based assays to measure cytokine production by a specific suppression; selection of the optimal set of markers for the subset may need to be fine-tuned such as by the use identification of activated T cells; and improved ability to of subset-depleted or subset-enriched PBMC preparations. differentiate Treg and CD4+ Teff responses (one schema has Another potential issue related to T cell assays is that the been offered by the HIPC consortium [146]). Additional concentration of whole protein or peptide in the cell culture improvements require establishing clear parameters that define may need to be titrated relative to the number of T cells. The memory versus naïve T cell populations, influence due to optimal concentration of protein required for proper evaluation bystander cells, and standard methods for PBMC harvest, of immune responses in vitro may be non-physiologic due to a preparation, and storage. Concurrent with efforts to standard- limited antigen-presenting population and/or co-stimulation. T ize, in vitro immunogenicity screening assays are being cell assays also require support with homeostatic cytokines incorporated into the preclinical pipeline by a number of drug (IL-2, IL-7, IL-15, etc.) to reduce bystander effects [40].In developers. addition, optimization of T cell concentrations may be required [143]. 3.7. Mouse models of in vivo Td immunogenicity of While “irrelevant” cells secreting cytokines in vitro may contribute to overestimation of the frequency of activated human therapeutic proteins T cells, their presence also contributes to the overall level of activation of cells in the assay. Cells such as NK cells, Important advances in understanding MHC restriction, map- CD8+ T cells, and may play a supportive role in ping of epitope recognition for murine epitopes presented by these cultures [144], thus their removal can modify the murine MHC, and T cell function have been achieved with in outcome of a truly representative immune response. vivo mouse studies. However, when we turn to in vivo mouse Finding the right balance between minimization of irrele- studies for prediction and validation of Td immunogenicity for vant immune responses and support for the in vitro immune clinically relevant proteins, there are two major limitations response is one of the major obstacles to widespread that must be taken into consideration. The first is that human adaptation of in vitro T cell assays for pre-clinical screening and murine proteins are not identical, thus administration of of protein therapeutics. protein therapeutics to mice may result in responses to Viable and functional T cells are also required for the components of the protein that are foreign to mice; and assay; standardized and optimized procedures for handling second, murine MHC will present mouse, not human, T cell and storing whole blood are needed to ensure the accuracy epitopes. Since murine models provide a means of evaluating of subsequent assays. Furthermore, blood from naïve and immunogenicity and an important bridge to the clinic, a drug-exposed individuals can differ in the content of number of enhanced models have been developed. antigen-specific T effector and T regulatory cells. Based on the nature of response (in vivo primed vs. stimulated and 3.7.1. HLA transgenic mice recalled), stimulation methods and the amount of antigen The HLA transgenic lines are generated by incorporation of required for challenge can also differ. Naïve cells will specific human HLA genes into murine MHC class II-deficient require multiple in vitro stimulations to amplify detection, mice, producing a mouse strain that expresses human class II while antigen-specific recall responses can be elicited even HLA in the absence of mouse class II MHC [147,148]. Thus, with a single challenge. these mice process and present epitopes in the context of 544 V. Jawa et al. human HLA, and their T cells recognize epitopes presented by With regard to the HLA DR transgenic models, those peptides human HLA. They are most useful when directly comparing two predicted to be HLA ligands were only immunogenic when the proteins that are very similar (such as FVIII and versions of FVIII sequence contained mismatches between the human FVIII that have fewer epitopes or new glycosylations) [97].Adirect sequence used in the immunization and the corresponding correlation has been found between epitopes that elicit T cell sequence in the native mouse FVIII protein [97].Tobe responses in infected humans and those that induce T cell immunogenic, the mismatches had to occur within 9-mer responses in immunized HLA transgenic mice [149–152]. sequences that also contained HLA DR3 or DR4 binding motifs. HLA transgenic mice are now routinely used to test and When presented to the mouse T cell, these 9-mers would optimize (human) epitope-driven vaccines in preclinical appear “foreign” and thus stimulate T cell proliferation. studies [153–155]. For example, Hanke et al. mapped HIV Peptides that were predicted to bind HLA DR3 and/or DR4 but epitopes in transgenic mice, and then moved their DNA vaccine did not stimulate immune responses were found to contain no through abbreviated primate studies after proving that the one mouse/human sequence mismatches. non-human primate epitope engineered into the vaccine In summary, identification of T cell epitopes and stimulated T cells [156]. development of de-immunized versions by targeted se- The formation of anti-FVIII antibodies, also known as quence modification can lower HLA binding and proliferation inhibitors, is a major obstacle to FVIII gene replacement responses, but the process has the potential to impact therapy in hemophilia A patients. After intravenous adminis- protein function tration of FVIII, the immune response mounted is dependent + on CD4 T helper cells, as has been demonstrated by numerous 3.7.2. Humanized mouse models – studies in mice and humans [157 162].Morecurrently, “Humanized” mice engrafted with a functional human – interference with T B cell interactions in hemophilic mice immune system are now being used to study human was shown to reduce inhibitor formation [159,160,162]. hematopoiesis, immunity, regeneration, stem cell function, AnexampleoftheuseofHLAtransgenicmiceforevaluating cancer, and human-specific infectious agents. Immunocom- the application of immunogenicity prediction tools towards promised SCID/NOD/γ chain−/− or RAG2−/−/γ chain−/− the goal of deimmunizing a therapeutic protein was recently mice, utilized as recipients to facilitate acceptance of published in Clinical Immunology [97]. In silico tools were human tissue, are engrafted with functional human hema- utilized to predict immunogenic peptides within the C2 domain topoietic stem cells (CD34+), liver, and thymus [165]. The of FVIII. Changes to amino acids in positions predicted to be result is a cohort of mice in which human myeloid and important for binding to the HLA DR3 MHC class II pocket were lymphoid lineages are reconstituted from a single human – modified with the intent of disrupting peptide MHC binding. donor, and the interactions of these cells in a complex The same predictive tools were reapplied to assess the binding biological environment can be studied. XenoMouse® (de- potential of the modified peptides. This process was reiterated scribed below), in addition to the humanized mice such until the predicted binding to HLA DR3 was reduced. as NOD/Shi-scid/IL-2Rγnull (NOG), NOD scid IL2 receptor The de-novo immunogenicity of these modified peptides was gamma chain knockout mice (NSG), bone marrow, liver, tested in hemophilic E16 mice (H-2b; [163]) and in HLA-DR3 thymus transplanted mouse (BLT), and bone marrow transgenic mice [164]. The initial immunogenicity study transplanted mouse (BMT) have all been used as animal results, in which mice were immunized with the unmodified models to evaluate human immune responses [165–168]. original (ORG) epitopes predicted by in silico analysis of the C2 domain, were consistent with predicted responses for either H-2b- or HLA-DR3-expressing mice. These two mouse strains 3.8. Strengths and limitations of mouse models of in were crossed to produce E16xDR3 mice, in which an immune vivo Td immunogenicity response of intermediate magnitude was observed. Specifi- cally, the immunogenicity of epitopes derived from FVIII in the These evolving mouse models could provide functional and E16xDR3 mice was consistent with the absence of tolerance testable elements of the innate and adaptive human immune induction to this sequence (the E16 mice do not express system without putting patients at risk [169]. The species full-length FVIII) and the presence of MHC (I-Ab) binding motifs specificity of a number of cells and molecules critical for a in the sequence [97]. In proliferation assays, modified epitope fully functional immune system remains a limitation in these peptides were less antigenic than ORG peptides. In general, models. For example, in HLA transgenic mice, the T cell lower antigenicity was observed for those peptides that had repertoire will be shaped by epitopes derived from mouse two rather than just one amino acid substitution. Similarly, in proteins presented by a single human HLA allele to mouse terms of de novo immunogenicity, the more mutations, the TCR. This confounds the application of HLA transgenic mice lower the observed proliferative response in general. for determination of immunogenicity of a human protein In addition to providing an example of how in silico therapeutic. And while humanized (SCID/Hu) mouse models tools can be applied early in the development process to are improving, certain aspects of a fully functional immune mitigate immunogenicity risk, this study also highlights response relevant to immunogenicity prediction and mitiga- certain limitations of available mouse models for risk tion are lacking, such as the ability to elicit the complete assessment. While certain epitopes are predicted to bind spectrum of B cell antibody responses or the ability to promiscuously to both human HLA DR3 and I-Ab MHC, others proteolytically process antigens in a way that recapitulates demonstrate greater restriction by either the human or the what has been observed for human endosomal/lysosomal mouse MHC. Thus the lack of observed immunogenicity may proteases. be attributed to the absence of a relevant MHC expressed in The XenoMouse® model has been of particular interest, as it the mouse model. is transgenic for nearly the complete human immunoglobulin Td immunogenicity of biologics 545 locus, thus is tolerant for human IgG2/kλ antibodies, but is showed relatively high antibody concentration with lower, deficient for mouse IgH and Igk chains. The human-like humoral but measurable, SFCs in vitro. Immune responses (both in immune response in XenoMouse® is restricted by mouse MHC vivo antibody and in vitro T cell) to different regions of and T cell help but is not as robust as in wild type mice, the protein correlated with in silico predictions. The potentially due to inefficient signal transduction and isotype carboxy-terminal region of the FPX peptide showed the switching mediated by accessory factorsthatarenecessaryfor highest MHC binding score in the context of the DRB1*0701 B cell maturation. Hence the utility of such a model to study allele; T cell and antibody responses to this fragment were immune responses to human proteins remains somewhat observed in vitro and in vivo, respectively, for individuals limited. possessing that allele. In contrast, the DRB1*0301 allele had Clearly, more work is required to develop these mouse very low MHC binding scores, and patients who possessed models to accurately reflect human immune responses to DRB1*0301 but not any of the other higher binding alleles protein therapeutics before they can become accurate, useful, demonstrated low responses in ELISpot and no evidence of and routine components of a Td immunogenicity screening an antibody response to the protein therapeutic. The program. immunogenicity of the FPX fragments and the association between clinical results and the HLA class II alleles were supported in the naïve blood T cell assays in further 4. Applied Td immunogenicity prediction studies performed by Jawa et al. [40,54] the reactivity of naïve (pre-exposure) PBMC to FPX1 was associated with 4.1. In silico prediction supported by subsequent therapeutic-induced antibody responses observed in the clinical data clinic as well as with expression of specific HLA class II alleles that were predicted in silico to present FPX1-derived Koren et al. [54] demonstrated a correlation between the in epitopes. silico evaluation of T helper epitope content of a protein This case study illustrates several important princi- therapeutic and its observed immunogenicity when adminis- ples regarding the immunogenicity assessment of protein tered to human subjects in a clinical trial. The therapeutic therapeutics: protein of interest was a recombinant Fc fusion protein (FPX) consisting of human germ line Fc γ fragment with two identical, (i) Clinical correlation: Clinical incidence of high immuno- biologically active, 24-amino-acid peptides attached to the genicity to FPX1 from exposed donors was retrospec- amino terminal end of the Fc fragment. In the in silico analysis, tively associated with the in silico immunogenicity the carboxy terminal region of the peptide scored high for predictions. binding to five of eight common HLA molecules, suggesting that (ii) Promiscuous epitopes: This study demonstrated that this peptide had the potential to be presented by five different immunodominance was associated with clusters of HLA molecules to T cells. Moreover, the C-terminus peptides epitopes within the sequence of the FPX1 peptide. were associated with a cluster of overlapping 9-mers that could The clustering of epitopes was associated with greater bind across several HLA DR alleles. immunogenicity as measured by a high incidence of The antibody response to FPX was consistent with the high binding and neutralizing antibodies to FPX fusion immunogenic potential predicted in silico. A single subcutane- protein. Moreover, due to the clustering, the peptide ous or intravenous administration of the protein resulted in was more promiscuous and was able to bind across high-affinity binding antibodies in 40% and 33% of total several HLA DR alleles. This was validated when ADA individuals, respectively (Table 3). T cell-mediated recall positive subjects were observed to express high-binding responses to the therapeutic protein were also assessed alleles. in vitro for donors exhibiting a strong humoral response (iii) Antibody–HLA correlation: The responders with high in vivo. In vitro PBMC activation by the FPX peptide, and the antibody titers expressed HLA-DR alleles that had been amino-terminal and carboxy-terminal fragments thereof, was predicted by the in silico algorithm to be the best measured as a function of the number of IFN-γ and IL-4 epitope binders. spot-forming cells (SFC) in a standard ELISpot assay. The (iv) In silico–T cell assay correlation: The C-terminal region antibody data suggested a strong T cell-driven response, which of the FPX1 peptide elicited T cell responses in PBMC from was corroborated by the in vitro cytokine responses observed in FPX1-exposed, antibody-positive donors, supporting a PBMC culture. Thus, the in silico prediction of immunogenic correlation between in silico prediction and observed T helper cell epitope(s) within the carboxy-terminal region of clinical immunogenicity. the FPX peptide correlated with the in vitro T cell assays and (v) Correlation to in vitro naïve response: The predicted the in vivo antibody responses. immunodominant regions of the peptide were able to HLA typing confirmed the predicted binding promiscuity elicit response from naïve PBMC with the HLA DR of the carboxy-terminal epitope(s), as antibody-positive predicted in silico. subjects possessed all of the eight most common HLA alleles. The magnitude of the immune response also appeared to correlate with the HLA haplotype and with the best 4.2. Clinical link between MHC class II haplotype and carboxy-terminal peptide binding scores. One subject who IFN-β immunogenicity [53] possessed the DRB1*0701 allele had the highest antibody concentration as well as the highest number of IFN-γ and A similar association between the HLA DRB1*0701 allele and IL-4 SFCs, as was predicted based on the in silico EpiMatrix a strong antibody response to recombinant beta-interferon results. Another subject who had DRB1*0701 allele also was observed by Barbosa et al. The IFN-β epitopes were 546 V. Jawa et al. identified using a peptide library and a peptide binding assay illustrates some of the limitations and lessons learned in with B cell lines expressing this allele. Peptides were the interpretation of in vitro studies: synthesized as overlapping 17-mers covering the entire sequence of IFN-β and combined into 10-peptide pools. (i) The overlapping peptide approach may lead to motif Recall responses from subjects with multiple sclerosis truncation: The approach of using overlapping pep- receiving Type I IFN-β therapy were assessed in vitro by tides that are then pooled for in vitro analysis may measuring T cell IFN-γ ELISpots (Fig. 1). The PBMC from make it difficult to define a specific epitope or epitope antibody-positive subjects (haplotype DRB1701/DQA10201) cluster, particularly if limited PBMC samples do not had high levels of IFN-γ-secreting T cells in the presence of accommodate de-convolution of these peptide pools two of the 10 peptide pools tested, while antibody-negative (T cell assays). subjects showed no response. Depending on the availability (ii) HLA skewing: Natural limitations on the number of of cells, peptide pools could be de-convoluted to identify subjects that are exposed to the drug or HLA-association the minimal epitope and the restricting HLA class II allele. of the disease itself may lead to skewed representation of One limitation of such an approach is the difficulty of HLA alleles, thus some associations between HLA and defining the true immunogenic epitope, and consequently clinical immunogenicity may be overlooked. the most relevant HLA-DR motifs, among the overlapping (iii) Combining methods may lead to new insights:Careful peptides, as is illustrated in Fig. 1. comparison of the HLA-binding and T cell assay outcomes Stickler et al. [170] observed that DRB1*1501/DQB1*0602 may uncover linkages that can better explain the results; haplotypes are associated with a high rate of reactivity limiting immunogenicity screening to a single approach towards human IFN-β in the naïve T cell assay format. may lead to misinterpretations. Accordingly, the C-terminal IFN-β peptide 147–161 was eluted from HLA-DR molecules after IFN-β exposure of human DC expressing the DRB1*1501 allele (H. Kropshofer, 5. Mitigation of T cell-dependent immunogenicity personal communication). Due to limitations in the number of subjects tested, a clinical correlation with the HLA- DRB1*1501 allele could not be detected in the Barbosa et al. Computer algorithms, epitope databases, and improved study [53]. Thus the Barbosa et al. publication nicely statistical methods have given researchers superior tools

A)

B)

Figure 1 Association of HLA with patient immune responses. Pools of overlapping 17mers derived from IFN-β were tested in ELISpot assays using PBMC from multiple sclerosis patients with antibody response to beta-interferon therapy (Barbosa et al., Clinical Immunology 2006). Pools 9 and 10 stimulated significant IFN-γ responses in patients sharing a single HLA allele, DRB1*0701, thus the minimal epitope for this allele was determined to comprise the residues shared between the two stimulatory pools. However, immunoinformatic analysis using the EpiMatrix algorithm reveals three distinct 9-mer binding motifs for DRB1*0701 in these peptides, none of which are fully represented in the region of overlap between the two peptide pools. Td immunogenicity of biologics 547 for assessing the potential of T cell immunogenicity of of course, clinical testing is the only way to “validate” any of therapeutic proteins [42]. As a result, a number of different the methods discussed in this review. Retrospective studies may approaches to mitigate Td immunogenicity of therapeutic be the only approach to testing the concept, but without a proteins are now under consideration. These approaches head-to-head comparison of the original and deimmunized include: direct modification of the therapeutic protein by proteins (FVIII, for example) in a clinical setting, all else will be pegylation and/or glycosylation to mask “immunogenic epi- correlative. topes”, thereby reducing recognition by the immune system; modification of HLA class II anchor residues of immunodominant 5.2. Tolerization epitopes to disrupt presentation; and application of strategies that tolerize the immune system to the therapeutic protein. Active interference with T cell responses to protein The primary focus of this review is on methods for predicting therapeutics by inducing tolerance to the drug is an and measuring Td immunogenicity; hence, a thorough discus- approach that has garnered significant interest in the last sion of methods for mitigating immunogenicity is beyond its few years. Tolerance can be induced with non-depleting scope. However, it is important here to link our evolving ability anti-CD4 antibodies [181,182]. IVIG has also been used to to identify contributors to a therapeutic protein's immunoge- induce tolerance in solid organ transplant [183], to reduce nicity with our ability to modify that feature to mitigate FVIII inhibitors [184–189], and to inhibit ADA in Pompe unwanted immune responses. Currently, tools to predict T cell patients undergoing Myozyme treatment [190]. Tolerance- epitopes can be applied to remove T cell epitopes. Indeed, a inducing protocols combining Rituxan, methotrexate and method for deimmunization (protein/sequence re-engineering) IVIG that have been effectively applied to mitigate existing has been introduced, which involves the elimination of anti-therapeutic responses in the context of life-saving predicted T cell epitopes or a reduction in the total number enzyme replacement therapy may provide an immediate of T cell epitopes. The approach has been described in detail solution for these patients [68,190,191]. However, there are – in a number of publications [41,171 175], and readers are new opportunities opening up along the developmental referred to those articles for details on the methodology. pipeline to build next-generation protein therapeutics to Efforts to render effector T cells non-responsive through the which patients are more tolerant. Towards this goal, De actions of immunosuppressive drugs or induction of Treg cells Groot et al. have identified a set of natural, human are evolving. Finally, the ability to identify drug-induced or regulatory T cell epitopes (“Tregitopes”) present in the Fc drug-responsive T cells may in the future bring opportunities and Fab domains of IgG that have also been shown to induce to specifically deplete them. tolerance to co-administered proteins; these Tregitopes may be responsible for tolerance to idiotypic epitopes. When incubated with PBMC in vitro, Tregitopes specifically 5.1. Deimmunization activate CD4+ T cells, increase CD25/Foxp3 expression, and increase expression of regulatory cytokines and chemokines The first published attempt to deimmunize a protein [41]. Administration of Tregitopes with protein antigens in involved the introduction of alanine substitutions at the vivo inhibits T cell proliferation and effector cytokine MHC anchoring residues Y73, K74, R77, E80, and D82 of expression, and induces antigen-specific adaptive tolerance staphylokinase, alone or in combination. These modifica- induction. Methods for co-administering Tregitopes with tions to the native protein were subsequently shown to protein therapeutics are currently under development reduce or eliminate a T cell response and clinical immuno- [192–195]. genicity [176]; this was followed by a number of additional – deimmunization studies [29,41,171 175,177,178]. Ongoing 6. Considerations: assessing immunogenicity of efforts include deimmunization of (1) botulinum neurotoxin type A, (2) lysostaphin, and (3) Factor VIII [97], as well as therapeutic proteins (4) removing T regulatory epitopes from a (anti-DEC205) for vaccine applications. Epitope modification Given the potential impact of immunogenicity on the safety, has also been applied to several other therapeutic proteins efficacy, and utility of a final therapeutic protein drug in studies performed by researchers at Biovation [179], product, there is increasing interest in developing ap- Epimmune [52],Genencor[180], and elsewhere, using a proaches that will help to establish an accurate immunoge- range of methodologies. nicity profile for a protein therapeutic. Immunogenicity Clearly, modification of the sequence of a protein can have predictions provide the greatest opportunity, and have the an impact on the secondary or tertiary structure. Thus the greatest benefit, early during selection of a lead molecule. utility of this approach depends on the location and extent of Introduction of changes to the structure of a lead molecule amino acid changes and the impact of those changes on the could be prohibitively expensive with respect to time and pharmacological activity of the modified protein. On a case-by- resources necessary to repeat prior development and safety case basis, an immunoinformatics-driven and cell-based ap- studies. We suggest consideration of a multi-step approach proach to modifying protein sequences may eventually lead to such as the following (Fig. 2) to evaluate the Td immunoge- less immunogenic proteins while preserving their therapeutic nicity potential of a therapeutic protein: potential. Several de-immunized therapeutic proteins are in clinical use [176]; however, in the majority of instances, the Step 1 HLA binding prediction in silico and confirmatory immunogenic potential of the parent molecule was determined assessment in vitro. A high-throughput in silico by the presence of T cell epitopes and not through clinical epitope-mapping algorithm applicable across com- testing. This is the key limitation of all predictive methods, and mon HLA class II haplotypes could identify areas of 548 V. Jawa et al.

In Silico In Vitro In Vivo

Screen multiple Examine: therapeutic Test PBMC • Non-sequence- Human candidates driven Responses immunogenicity SCID Mouse • Natural antigen Model Perform T processing and Rank for Vary Ratio by DCs immunogenicity Cell Assays Proceed to based on with of T cells to • Peptide/MHC Either Drug proteins / stability predicted dendritic Development epitope content target cells • T cell activation epitopes thresholds HLA • Post- Transgenic translational Mouse Modify sequences Investigate modification Model to reduce artificial LN • Formulation- immunogenicity induced changes

Figure 2 Example roadmap for immunogenicity prediction. A potential step-wise approach to pre-clinical immunogenicity testing is depicted here. Step 1 would consist of the in silico screening of linear sequences from multiple therapeutic candidates for T cell epitopes and clusters. At the conclusion of Step 1, therapeutic candidates would be rated for immunogenic risk. Step 2 would be an in vitro evaluation of immunogenicity in a series of T cell assays. This phase could include examination of antigen processing and presentation, post-translational modifications, and different formulations. Subsequent Step 3 immunogenicity testing of therapeutic candidates may be carried out in vivo in established “humanized” animal models such as the Hu-SCID and HLA-transgenic mice. The culmination of this immunogenicity prediction strategy would be advancement into the drug development pipeline.

putative epitope clusters within a larger protein and screening may enable the evaluation of processing- allow drug candidates to be ranked in order of their associated changes, such as posttranslational modi- potential for immunogenicity. However, in silico fications and misfolding. Where target-mediated, screening should never be viewed as a stand-alone agonist effects, or formulation changes of an immuno- technology. The in silico results could be carried modulatroy therapeutic are a concern, in vitro studies forward by having the epitopes identified synthe- can be performed in tandem with in silico studies. sized for testing in an in vitro peptide-binding assay. From this analysis, epitopes could be prioritized by Immunogenicity assessment is an important adjunct to strength of binding (IC50) and by the breadth of HLA preclinical studies that may facilitate the identification and to which they bind. To strengthen the predictive selection of the best candidates to bring towards the clinic. The value of in silico tools and HLA binding assays that do stepwise approach to preclinical Td immunogenicity assessment not take into account antigen processing, editing, or proposed here progresses from higher-throughput, lower-cost presentation, cell-based approaches as described methods to narrow the search space, to lower-throughput, above [103] may also be applied. higher-cost methods to screen for T cell reactivity. At this point, Step 2 Stimulation of T cell activity in vitro. The focused set a combination of these experimental results, experience of epitopes or portions of a larger protein identified in with the protein of interest, and the stage of the development Step 1 could be further assessed by in vitro human T process may together prompt an effort to either proceed cell assays to study immune responsiveness to this into in vivo animal experiments and further development protein in Step 2. Knowledge gained from the in silico (Fig. 2, Step 3), or perhaps to reengineer the protein to lower analysis can be applied here to demonstrate specificity immunogenicity. of the response. Proliferation assays are sensitive, low in cost, and can be designed to assess phenotype of the responder cells. ELISpot assays and multiplex cytokine 7. Conclusion bead arrays are moderate in cost and more sensitive; intracellular cytokine staining methods have the In response to concerns about the potential side effects of highest sensitivity but can be technically challenging anti-drug immune responses, regulatory bodies such as the and higher in cost. It is recommended to utilize at least FDA and the EMA have begun to request that each protein two independent readouts of T cell activity to ensure a therapeutic be accompanied by an immunogenicity risk balance between specificity and sensitivity. Results assessment [196,197]. For example, the recent EMA guid- from Step 2 should strengthen the findings of the ance mentions “predictive immunogenicity” as an approach in silico prediction from Step 1. In addition, in vitro sponsors could consider in their preclinical studies [17].In Td immunogenicity of biologics 549 this context, the historical focus has been on measurement conflict of interest related to their relationship with EpiVax of antibody responses as the read-out for immunogenicity, and attest that the work contained in this report is free of any supported by the obvious consequences of ADA responses on bias that might be associated with the commercial goals of the protein therapeutic pharmacokinetics and efficacy. Td company. immunogenicity assessment has been considered by many drug developers to be an “upstream” activity, associated References with the lead candidate optimization/selection process. Given the contribution of T cell responses to the develop- [1] D.S. Dimitrov, Therapeutic proteins, Methods Mol. Biol. 899 ment of a detrimental ADA response and the emerging suite (2012) 1–26. of tools for predicting Td immunogenicity, this focus is [2] W.R. Strohl, D.M. Knight, Discovery and development of shifting in some instances. For example, mitigating immu- biopharmaceuticals: current issues, Curr. Opin. Biotechnol. nogenicity by modification of T cell responses, with careful 20 (2009) 668–672. attention to potential Treg epitopes, is now mentioned in [3] G. Walsh, Biopharmaceutical benchmarks 2010, Nat. Biotechnol. the recently released FDA draft guidance entitled “Immu- 28 (2010) 917–924. nogenicity Assessment for Therapeutic Protein Products” [4] M.D. Barbosa, Immunogenicity of biotherapeutics in the [120]. context of developing biosimilars and biobetters, Drug Discov. Today 16 (2011) 345–353. The drug discovery and early development phases present [5] K.U. Eckardt, N. Casadevall, Pure red-cell aplasia due to prime opportunities to extract the greatest value from anti-erythropoietin antibodies, Nephrol. Dial. Transplant. 18 immunogenicity prediction tools. The cost of modifying or, in (2003) 865–869. extreme cases, abandoning drugs that are near or have reached [6] S.M. Tatarewicz, X. Wei, S. Gupta, D. Masterman, S.J. clinical trial stage can be quite significant [6]. Thus applying Td Swanson, M.S. Moxness, Development of a maturing T-cell- immunogenicity screening earlier in the drug-development mediated immune response in patients with idiopathic pathway may be an important adjunct to preclinical studies in Parkinson's disease receiving r-metHuGDNF via continuous some circumstances. intraputaminal infusion, J. Clin. Immunol. 27 (2007) 620–627. At present, drug developers are applying a range of [7] V. Jawa, M. Hokom, Z. Hu, N. El-Abaadi, Y. Zhuang, D. Berger, strategies and assays to assess and modulate humoral and/or S. Gupta, S.J. Swanson, N. Chirmule, Assessment of immuno- genicity of romiplostim in clinical studies with ITP subjects, cell-mediated immune responses to protein therapeutic drugs Ann. Hematol. 89 (2010) S75–S85. at both the preclinical and clinical phases of development. [8] G. Shankar, C. Pendley, K.E. Stein, A risk-based bioanalytical Guidelines for standardizing immunogenicity testing of protein strategy for the assessment of antibody immune responses therapeutics across the industry are emerging from these against biological drugs, Nat. Biotechnol. 25 (2007) 555–561. activities and parallel assay standardization and validation [9] M. Sauerborn, E. van de Vosse, D. Delawi, J.T. van Dissel, V. efforts in the vaccine discovery field. Approaches for reducing Brinks, H. Schellekens, Natural antibodies against bone biologic protein immunogenicity are also proceeding apace; a morphogenic proteins and interferons in healthy donors and number of drugs that have been de-immunized or tolerized are in patients with infections linked to type-1 cytokine re- now being introduced into the clinic. Additional methods, such sponses, J. Interferon Cytokine Res. 31 (2011) 661–669. as tolerization using cytotoxic and immunosuppressive drugs, [10] N.Casadevall,J.Nataf,B.Viron,A.Kolta,J.J.Kiladjian,P. Martin-Dupont, P. Michaud, T. Papo, V. Ugo, I. Teyssandier, B. are currently in use in high-risk enzyme replacement therapy Varet, P. Mayeux, Pure red-cell aplasia and antierythropoietin patients [66,68,191,198]. antibodies in patients treated with recombinant erythropoietin, A wide range of Td immunogenicity screening methods N. Engl. J. Med. 346 (2002) 469–475. examining different aspects of the process by which a [11] J. Li, C. Yang, Y. Xia, A. Bertino, J. Glaspy, M. Roberts, D.J. protein therapeutic may trigger an immune response has Kuter, Thrombocytopenia caused by the development of been presented here. No single method has emerged as a antibodies to thrombopoietin, Blood 98 (2001) 3241–3248. definitive tool for determining whether or not a protein [12] F. Baert, M. Noman, S. Vermeire, G. Van Assche, D.H.G., A. therapeutic will elicit a detrimental immune response in Carbonez, P. Rutgeerts, Influence of immunogenicity on the patients, and given the complexity of the immune system, long-term efficacy of infliximab in Crohn's disease, N. Engl. J. – a singular solution may not be realistic. Rather, the field Med. 348 (2003) 601 608. [13] C.H. Chung, B. Mirakhur, E. Chan, Q.T. Le, J. Berlin, M. is evolving to apply strategic combinations of multiple methods Morse, B.A. Murphy, S.M. Satinover, J. Hosen, D. Mauro, R.J. to most closely predict and mitigate immunogenicity risk. Td Slebos, Q. Zhou, D. Gold, T. Hatley, D.J. Hicklin, T.A. immunogenicity screening is a rapidly advancing science that Platts-Mills, Cetuximab-induced anaphylaxis and IgE specific could eventually impact drug development, reducing risks to for galactose-alpha-1,3-galactose, N. Engl. J. Med. 358 (2008) patients and costs to industry. As more preclinical immunoge- 1109–1117. nicity testing is performed and clinical correlations become [14] S.K. Singh, Impact of product-related factors on immunoge- available, the accuracy of preclinical immunogenicity screening nicity of biotherapeutics, J. Pharm. Sci. 100 (2011) 354–387. methods and their utility to industry are bound to improve. [15] A.H. Gutierrez, L. Moise, A.S. De Groot, Of [hamsters] and men: a new perspective on host cell proteins, Hum. Vaccin. Immunother. 8 (2012). Conflict of interest statement [16] K. Lewis, Inspiration Biopharmaceuticals Announces Clinical Hold of Clinical Trials Evaluating IB1001 for the Treatment and Prevention of Bleeding in Hemophilia B, Inspiration Anne S. De Groot is a senior officer/majority shareholder and Biopharmaceuticals, 2012. Leslie Cousens is an employee at EpiVax, Inc., a privately- [17] In: E.M. Agency (Ed.), Guideline on immunogenicity assess- owned immunotherapeutics company located in Providence ment of biotechnology-derived therapeutic proteins EMEA/ RI. These authors acknowledge that there is a potential CHMP/BMWP/14327/2007, 2007. 550 V. Jawa et al.

[18] B.Monzavi-Karbassi,S.Shamloo,M.Kieber-Emmons,F. Zuckerman, E. Koren, Recommendations for the validation of Jousheghany, P. Luo, K.Y. Lin, G. Cunto-Amesty, D.B. Weiner, immunoassays used for detection of host antibodies against T. Kieber-Emmons, Priming characteristics of peptide biotechnology products, J. Pharm. Biomed. Anal. 48 (2008) of carbohydrate antigens, Vaccine 21 (2003) 753–760. 1267–1281. [19] A. Beck, E. Wagner-Rousset, M.C. Bussat, M. Lokteff, C. [33] In: U.F.C. CBER (Ed.), Assay development for immunogenicity Klinguer-Hamour, J.F. Haeuw, L. Goetsch, T. Wurch, A. Van testing of therapeutic proteins, 2009. Dorsselaer, N. Corvaia, Trends in glycosylation, glycoanalysis [34] M. Stickler, N. Rochanayon, O.J. Razo, J. Mucha, W. Gebel, N. and glycoengineering of therapeutic antibodies and Fc-fusion Faravashi, R. Chin, S. Holmes, F.A. Harding, An in vitro human proteins, Curr. Pharm. Biotechnol. 9 (2008) 482–501. cell-based assay to rank the relative immunogenicity of [20] R. Mueller, A. Karle, A. Vogt, H. Kropshofer, A. Ross, K. Maeder, proteins, Toxicol. Sci. 77 (2004) 280–289. H.C. Mahler, Evaluation of the immuno-stimulatory potential of [35] A.S. De Groot, D.W. Scott, Immunogenicity of protein stopper extractables and leachables by using dendritic cells as therapeutics, Trends Immunol. 28 (2007) 482–490. readout, J. Pharm. Sci. 98 (2009) 3548–3561. [36] A.S. De Groot, J. McMurry, L. Moise, Prediction of immuno- [21] A.S. Rosenberg, Effects of protein aggregates: an immuno- genicity: in silico paradigms, ex vivo and in vivo correlates, logic perspective, AAPS J. 8 (2006) E501–E507. Curr. Opin. Pharmacol. 8 (2008) 620–626. [22] I.C. Buttel, P. Chamberlain, Y. Chowers, F. Ehmann, A. [37] L.C. Perry, T.D. Jones, M.P. Baker, New approaches to Greinacher, R. Jefferis, D. Kramer, H. Kropshofer, P. Lloyd, prediction of immune responses to therapeutic proteins A. Lubiniecki, R. Krause, A. Mire-Sluis, T. Platts-Mills, J.A. during preclinical development, Drugs R D 9 (2008) 385–396. Ragheb, B.M. Reipert, H. Schellekens, R. Seitz, P. Stas, M. [38] P. Stas, I. Lasters, Strategies for preclinical immunoge- Subramanyam, R. Thorpe, J.H. Trouvin, M. Weise, J. nicity assessment of protein therapeutics, IDrugs 12 (2009) Windisch, C.K. Schneider, Taking immunogenicity assessment 169–173. of therapeutic proteins to the next level, Biologicals 39 (2011) [39] F.A. Harding, M.M. Stickler, J. Razo, R.B. DuBridge, The 100–109. immunogenicity of humanized and fully human antibodies: [23] M.K. Joubert, M. Hokom, C. Eakin, L. Zhou, M. Deshpande, residual immunogenicity resides in the CDR regions, MAbs 2 M.P. Baker, T.J. Goletz, B.A. Kerwin, N. Chirmule, L.O. Narhi, (2010) 256–265. V. Jawa, Highly aggregated antibody therapeutics can [40] D. Wullner, L. Zhou, E. Bramhall, A. Kuck, T.J. Goletz, S. enhance the in vitro innate and late-stage T-cell immune Swanson, N. Chirmule, V. Jawa, Considerations for optimiza- responses, J. Biol. Chem. 287 (2012) 25266–25279. tion and validation of an in vitro PBMC derived T cell assay for [24] M.K. Joubert, Q. Luo, Y. Nashed-Samuel, J. Wypych, L.O. immunogenicity prediction of biotherapeutics, Clin. Immunol. Narhi, Classification and characterization of therapeutic 137 (2010) 5–14. antibody aggregates, J. Biol. Chem. 286 (2011) 25118–25133. [41] A.S. De Groot, L. Moise, J.A. McMurry, E. Wambre, L. Van [25] D. Verthelyi, V. Wang, Trace levels of innate immune Overtvelt, P. Moingeon, D.W. Scott, W. Martin, Activation response modulating impurities (IIRMIs) synergize to break of natural regulatory T cells by IgG Fc-derived peptide tolerance to therapeutic proteins, PLoS One 5 (2010) e15252. “Tregitopes”, Blood 112 (2008) 3303–3311. [26] M.M. van Beers, W. Jiskoot, H. Schellekens, On the role of [42] C.J. Bryson, T.D. Jones, M.P. Baker, Prediction of immuno- aggregates in the immunogenicity of recombinant human genicity of therapeutic proteins: validity of computational interferon beta in patients with multiple sclerosis, J. tools, BioDrugs 24 (2010) 1–8. Interferon Cytokine Res. 30 (2010) 767–775. [43] M.C. Honeyman, N.L. Stone, B.A. Falk, G. Nepom, L.C. [27] A.S. Rathore, H. Winkle, Quality by design for biopharmaceuticals, Harrison, Evidence for molecular mimicry between human T Nat. Biotechnol. 27 (2009) 26–34. cell epitopes in rotavirus and pancreatic islet autoantigens, [28] K. Praditpornsilpa, P. Kupatawintu, W. Mongkonsritagoon, O. J. Immunol. 184 (2010) 2204–2210. Supasyndh, S. Jootar, T. Intarakumthornchai, C. Pongskul, W. [44] M.F. Cusick, J.E. Libbey, R.S. Fujinami, Molecular mimicry as Prasithsirikul, B. Achavanuntakul, P. Ruangkarnchanasetr, S. a mechanism of autoimmune disease, Clin. Rev. Laohavinij, S. Eiam-Ong, The association of anti-r-HuEpo- Immunol. 42 (1) (2012) 102–111. associated pure red cell aplasia with HLA-DRB1*09-DQB1*0309, [45] R.F. Siliciano, A.D. Keegan, R.Z. Dintzis, H.M. Dintzis, H.S. Nephrol. Dial. Transplant. 24 (2009) 1545–1549. Shin, The interaction of nominal antigen with T cell antigen [29] A.R. Mire-Sluis, Y.C. Barrett, V. Devanarayan, E. Koren, H. receptors. I. Specific binding of multivalent nominal antigen Liu, M. Maia, T. Parish, G. Scott, G. Shankar, E. Shores, S.J. to cytolytic T cell clones, J. Immunol. 135 (1985) 906–914. Swanson, G. Taniguchi, D. Wierda, L.A. Zuckerman, Recom- [46] R.H. Zubler, Naive and memory B cells in T-cell-dependent and mendations for the design and optimization of immunoassays T-independent responses, Springer Semin. Immunopathol. 23 used in the detection of host antibodies against biotechnology (2001) 405–419. products, J. Immunol. Methods 289 (2004) 1–16. [47] C.C. Goodnow, C.G. Vinuesa, K.L. Randall, F. Mackay, R. Brink, [30] S. Gupta, S.R. Indelicato, V. Jethwa, T. Kawabata, M. Kelley, Control systems and decision making for antibody production, A.R. Mire-Sluis, S.M. Richards, B. Rup, E. Shores, S.J. Nat. Immunol. 11 (2010) 681–688. Swanson, E. Wakshull, Recommendations for the design, [48] M. Onda, Reducing the immunogenicity of protein therapeu- optimization, and qualification of cell-based assays used for tics, Curr. Drug Targets 10 (2009) 131–139. the detection of neutralizing antibody responses elicited to [49] T. Defrance, M. Taillardet, L. Genestier, T cell-independent B biological therapeutics, J. Immunol. Methods 321 (2007) 1–18. cell memory, Curr. Opin. Immunol. 23 (2011) 330–336. [31] E. Koren, H.W. Smith, E. Shores, G. Shankar, D. Finco-Kent, B. [50] J.M. Weaver, C.A. Lazarski, K.A. Richards, F.A. Chaves, S.A. Rup, Y.C. Barrett, V. Devanarayan, B. Gorovits, S. Gupta, T. Jenks, P.R. Menges, A.J. Sant, Immunodominance of CD4 T Parish, V. Quarmby, M. Moxness, S.J. Swanson, G. Taniguchi, cells to foreign antigens is peptide intrinsic and independent L.A. Zuckerman, C.C. Stebbins, A. Mire-Sluis, Recommenda- of molecular context: implications for vaccine design, J. tions on risk-based strategies for detection and characterization Immunol. 181 (2008) 3039–3048. of antibodies against biotechnology products, J. Immunol. [51] M. Moxness, S. Tatarewicz, D. Weeraratne, N. Murakami, D. Methods 333 (2008) 1–9. Wullner, D. Mytych, V. Jawa, E. Koren, S.J. Swanson, [32] G. Shankar, V. Devanarayan, L. Amaravadi, Y.C. Barrett, R. Immunogenicity testing by electrochemiluminescent detec- Bowsher, D. Finco-Kent, M. Fiscella, B. Gorovits, S. Kirschner, tion for antibodies directed against therapeutic human M. Moxness, T. Parish, V. Quarmby, H. Smith, W. Smith, L.A. monoclonal antibodies, Clin. Chem. 51 (2005) 1983–1985. Td immunogenicity of biologics 551

[52] S. Tangri, B.R. Mothe, J. Eisenbraun, J. Sidney, S. Southwood, peptides reveal highly conserved immunogenic flanking K.Briggs,J.Zinckgraf,P.Bilsel,M.Newman,R.Chesnut, region sequence preferences that reflect antigen processing C. Licalsi, A. Sette, Rationally engineered therapeutic rather than peptide–MHC interactions, J. Immunol. 166 proteins with reduced immunogenicity, J. Immunol. 174 (2005) (2001) 6720–6727. 3187–3196. [71] I.A. Doytchinova, V.A. Walshe, N.A. Jones, S.E. Gloster, P. [53] M.D. Barbosa, J. Vielmetter, S. Chu, D.D. Smith, J. Jacinto, Borrow, D.R. Flower, Coupling in silico and in vitro analysis of Clinical link between MHC class II haplotype and interferon- peptide–MHC binding: a bioinformatic approach enabling beta (IFN-beta) immunogenicity, Clin. Immunol. 118 (2006) prediction of superbinding peptides and anchorless epitopes, 42–50. J. Immunol. 172 (2004) 7495–7502. [54] E. Koren, A.S. De Groot, V. Jawa, K.D. Beck, T. Boone, D. [72] A.S. De Groot, W. Martin, Reducing risk, improving outcomes: Rivera, L. Li, D. Mytych, M. Koscec, D. Weeraratne, S. bioengineering less immunogenic protein therapeutics, Clin. Swanson, W. Martin, Clinical validation of the “in silico” Immunol. 131 (2009) 189–201. prediction of immunogenicity of a human recombinant [73] Y. El-Manzalawy, D. Dobbs, V. Honavar, Predicting MHC-II therapeutic protein, Clin. Immunol. 124 (2007) 26–32. binding affinity using multiple instance regression, IEEE/ACM [55] K.P. Pratt, A.R. Thompson, B-cell and T-cell epitopes Trans. Comput. Biol. Bioinform. 8 (4) (2011) 1067–1079. in anti-factor VIII immune responses, Clin. Rev. Allergy [74] A. Sette, S. Buus, E. Appella, J.A. Smith, R. Chesnut, C. Miles, Immunol. 37 (2009) 80–95. S.M. Colon, H.M. Grey, Prediction of major histocompatibility [56] C. Kamate, P.J. Lenting, H.M. van den Berg, T. Mutis, complex binding regions of protein antigens by sequence Depletion of CD4+/CD25high regulatory T cells may enhance pattern analysis, Proc. Natl. Acad. Sci. U. S. A. 86 (1989) or uncover factor VIII-specific T-cell responses in healthy 3296–3300. individuals, J. Thromb. Haemost. 5 (2007) 611–613. [75] A.S. De Groot, J. Rayner, W. Martin, Modelling the immuno- [57] S. Yarkoni, A. Kaminitz, Y. Sagiv, I. Yaniv, N. Askenasy, genicity of therapeutic proteins using T cell epitope mapping, Involvement of IL-2 in homeostasis of regulatory T cells: the Dev. Biol. (Basel) 112 (2003) 71–80. IL-2 cycle, Bioessays 30 (2008) 875–888. [76] A.S. De Groot, J.A. Berzofsky, From genome to vaccine—new [58] J.A. Bluestone, A.K. Abbas, Natural versus adaptive regula- immunoinformatics tools for vaccine design, Methods 34 tory T cells, Nat. Rev. Immunol. 3 (2003) 253–257. (2004) 425–428. [59] N.M. Tsuji, A. Kosaka, Oral tolerance: intestinal homeostasis [77] G.L. Zhang, N. Petrovsky, C.K. Kwoh, J.T. August, V. Brusic, and antigen-specific regulatory T cells, Trends Immunol. 29 PRED(TAP): a system for prediction of peptide binding to (2008) 532–540. the human transporter associated with antigen processing, [60] M.C. Eyerman, X. Zhang, L.J. Wysocki, T cell recognition Immunome Res. 2 (2006) 3. and tolerance of antibody diversity, J. Immunol. 157 (1996) [78] J.C. Tong, Z.H. Zhang, J.T. August, V. Brusic, T.W. Tan, S. 1037–1046. Ranganathan, In silico characterization of immunogenic [61] D.T. Umetsu, R.H. DeKruyff, The regulation of allergy and epitopes presented by HLA-Cw*0401, Immunome Res. 3 asthma, Immunol. Rev. 212 (2006) 238–255. (2007) 7. [62] M.R. Karlsson, J. Rugtveit, P. Brandtzaeg, -responsive [79] G.L. Zhang, A.M. Khan, K.N. Srinivasan, A. Heiny, K. Lee, C.K. CD4+CD25+ regulatory T cells in children who have outgrown Kwoh, J.T. August, V. Brusic, Hotspot Hunter: a computa- cow's milk allergy, J. Exp. Med. 199 (2004) 1679–1688. tional system for large-scale screening and selection of [63] A.R. Kendal, Y. Chen, F.S. Regateiro, J. Ma, E. Adams, S.P. candidate immunological hotspots in pathogen proteomes, Cobbold, S. Hori, H. Waldmann, Sustained suppression by BMC Bioinforma. 9 (Suppl. 1) (2008) S19. Foxp3+ regulatory T cells is vital for infectious transplanta- [80] D.R. Flower, K. Phadwal, I.K. Macdonald, P.V. Coveney, M.N. tion tolerance, J. Exp. Med. 208 (2011) 2043–2053. Davies, S. Wan, T-cell epitope prediction and immune [64] M. Congia, S. Patel, A.P. Cope, S. De Virgiliis, G. Sonderstrup, complex simulation using molecular dynamics: state of the T cell epitopes of insulin defined in HLA-DR4 transgenic mice art and persisting challenges, Immunome Res. 6 (Suppl. 2) are derived from preproinsulin and proinsulin, Proc. Natl. (2010) S4. Acad. Sci. U. S. A. 95 (1998) 3833–3838. [81] T. Sturniolo, E. Bono, J. Ding, L. Raddrizzani, O. Tuereci, U. [65] R.J. Farrell, M. Alsahli, Y.T. Jeen, K.R. Falchuk, M.A. Peppercorn, Sahin, M. Braxenthaler, F. Gallazzi, M.P. Protti, F. Sinigaglia, P. Michetti, Intravenous hydrocortisone premedication reduces J. Hammer, Generation of tissue-specific and promiscuous antibodies to infliximab in Crohn's disease: a randomized HLA ligand databases using DNA microarrays and virtual HLA controlled trial, Gastroenterology 124 (2003) 917–924. class II matrices, Nat. Biotechnol. 17 (1999) 555–561. [66] M. Rohrbach, A. Klein, A. Kohli-Wiesner, D. Veraguth, I. [82] S. Southwood, J. Sidney, A. Kondo, M.F. del Guercio, E. Appella, Scheer, C. Balmer, R. Lauener, M.R. Baumgartner, CRIM- S.Hoffman,R.T.Kubo,R.W.Chesnut,H.M.Grey,A.Sette, negative infantile Pompe disease: 42-month treatment Several common HLA-DR types share largely overlapping peptide outcome, J. Inherit. Metab. Dis. 33 (2010) 751–757. binding repertoires, J. Immunol. 160 (1998) 3363–3373. [67] E. Moran, F. Carbone, V. Augusti, F. Patrone, A. Ballestrero, [83] D. Valmori, F. Levy, E. Godefroy, L. Scotto, N.E. Souleimanian, J. A. Nencioni, Proteasome inhibitors as immunosuppressants: Karbach, V. Tosello, C.S. Hesdorffer, L.J. Old, E. Jager, M. biological rationale and clinical experience, Semin. Hematol. Ayyoub, Epitope clustering in regions undergoing efficient 49 (2012) 270–276. proteasomal processing defines immunodominant CTL regions of [68] A. Joseph, K. Neff, J. Richard, L. Gao, D. Bangari, M. Joly, K. a tumor antigen, Clin. Immunol. 122 (2007) 163–172. Culm-Merdek, R. Garman, J. Williams, S. Richards, M. Ruzek, [84] K.B.Bond,B.Sriwanthana,T.W.Hodge,A.S.DeGroot, Transient low-dose methotrexate induces tolerance to murine T.D.Mastro,N.L.Young,N.Promadej,J.D.Altman,K. anti- globulin and together they promote long-term Limpakarnjanarat, J.M. McNicholl, An HLA-directed mo- allograft survival, J. Immunol. 189 (2012) 732–743. lecular and bioinformatics approach identifies new HLA-A11 [69] Y. Kim, A. Sette, B. Peters, Applications for T-cell epitope HIV-1 subtype E cytotoxic T lymphocyte epitopes in HIV-1-infected queries and tools in the Immune Epitope Database and Thais, AIDS Res. Hum. Retroviruses 17 (2001) 703–717. Analysis Resource, J. Immunol. Methods 374 (1–2) (2011) [85] A.S. De Groot, A. Bosma, N. Chinai, J. Frost, B.M. Jesdale, 62–69. M.A. Gonzalez, W. Martin, C. Saint-Aubin, From genome to [70] A.J. Godkin, K.J. Smith, A. Willis, M.V. Tejada-Simon, J. vaccine: in silico predictions, ex vivo verification, Vaccine 19 Zhang, T. Elliott, A.V. Hill, Naturally processed HLA class II (2001) 4385–4395. 552 V. Jawa et al.

[86] Y. Dong, S. Demaria, X. Sun, F.R. Santori, B.M. Jesdale, A.S. De [103] H. Kropshofer, T. Singer, Overview of cell-based tools for Groot, W.N. Rom, Y. Bushkin, HLA-A2-restricted CD8+-cytotoxic- pre-clinical assessment of immunogenicity of biotherapeutics, J. T-cell responses to novel epitopes in Mycobacterium tuberculosis Immunotoxicol. 3 (2006) 131–136. superoxide dismutase, alanine dehydrogenase, and glutamine [104] T.A.Rohn,A.Reitz,A.Paschen,X.D.Nguyen,D.Schadendorf, synthetase, Infect. Immun. 72 (2004) 2412–2415. A.B. Vogt, H. Kropshofer, A novel strategy for the discovery [87] J. McMurry, H. Sbai, M.L. Gennaro, E.J. Carter, W. Martin, A.S. of MHC class II-restricted tumor antigens: identification of a De Groot, Analyzing Mycobacterium tuberculosis proteomes for melanotransferrin helper T-cell epitope, Cancer Res. 65 (2005) candidate vaccine epitopes, Tuberculosis (Edinb.) 85 (2005) 10068–10078. 95–105. [105] R.M. Chicz, R.G. Urban, Analysis of MHC-presented peptides: [88] O.A. Koita, D. Dabitao, I. Mahamadou, M. Tall, S. Dao, A. applications in autoimmunity and vaccine development, Tounkara, H. Guiteye, C. Noumsi, O. Thiero, M. Kone, D. Immunol. Today 15 (1994) 155–160. Rivera, J.A. McMurry, W. Martin, A.S. De Groot, Confirmation [106] C.A.Lazarski,F.A.Chaves,S.A.Jenks,S.Wu,K.A.Richards,J.M. of immunogenic consensus sequence HIV-1 T-cell epitopes in Weaver, A.J. Sant, The kinetic stability of MHC class II:peptide Bamako, Mali and Providence, Rhode Island, Hum. Vaccin. 2 complexes is a key parameter that dictates immunodominance, (2006) 119–128. Immunity 23 (2005) 29–40. [89] T. Cohen, L. Moise, M. Ardito, W. Martin, A.S. De Groot, A [107] K. Berer, G. Krishnamoorthy, Commensal gut flora and brain method for individualizing the prediction of immunogenicity autoimmunity: a love or hate affair? Acta Neuropathol. 123 of protein vaccines and biologic therapeutics: individualized (2012) 639–651. T cell epitope measure (iTEM), J. Biomed. Biotechnol. 2010 [108] R.M. Welsh, J.W. Che, M.A. Brehm, L.K. Selin, Heterolo- (2010). gous immunity between viruses, Immunol. Rev. 235 (2010) [90] L. Moise, R.M. Buller, J. Schriewer, J. Lee, S.E. Frey, D.B. 244–266. Weiner, W. Martin, A.S. De Groot, VennVax, a DNA-prime, [109] P. Wang, J. Sidney, C. Dow, B. Mothe, A. Sette, B. Peters, A peptide-boost multi-T-cell epitope poxvirus vaccine, induces systematic assessment of MHC class II peptide binding protective immunity against vaccinia infection by T cell predictions and evaluation of a consensus approach, PLoS response alone, Vaccine 29 (3) (2011) 501–511. Comput. Biol. 4 (2008) e1000048. [91] J.D. Ahlers, I.M. Belyakov, E.K. Thomas, J.A. Berzofsky, [110] L.A. Trouw, T.W. Huizinga, R.E. Toes, Autoimmunity in High-affinity T helper epitope induces complementary helper rheumatoid arthritis: different antigens—common principles, and APC polarization, increased CTL, and protection against Ann. Rheum. Dis. 72 (Suppl. 2) (2013) ii132–ii136. viral infection, J. Clin. Invest. 108 (2001) 1677–1685. [111] J.H. Kessler, W.E. Benckhuijsen, T. Mutis, C.J. Melief, S.H. [92] A.S. De Groot, W. Martin, From immunome to vaccine: van der Burg, J.W. Drijfhout, Competition-based cellular epitope mapping and vaccine design tools, Novartis Found. peptide binding assay for HLA class I, Curr. Protoc. Immunol. Symp. 254 (2003) 57–72(discussion 72-56, 98-101, 250-102). (2004)(Chapter 18, Unit 18 12). [93] A.S. De Groot, D.S. Rivera, J.A. McMurry, S. Buus, W. Martin, [112] J.H.Kessler,B.Mommaas,T.Mutis,I.Huijbers,D.Vissers,W.E. Identification of immunogenic HLA-B7 “Achilles' heel” epi- Benckhuijsen, G.M. Schreuder, R. Offringa, E. Goulmy, C.J. topes within highly conserved regions of HIV, Vaccine 26 Melief,S.H.vanderBurg,J.W.Drijfhout, Competition-based (2008) 3059–3071. cellular peptide binding assays for 13 prevalent HLA class I alleles [94] A.S. De Groot, M. Ardito, E.M. McClaine, L. Moise, W.D. using fluorescein-labeled synthetic peptides, Hum. Immunol. 64 Martin, Immunoinformatic comparison of T-cell epitopes (2003) 245–255. contained in novel swine-origin influenza A (H1N1) virus [113] A.C. Steere, W. Klitz, E.E. Drouin, B.A. Falk, W.W. Kwok, G.T. with epitopes in 2008–2009 conventional influenza vaccine, Nepom, L.A. Baxter-Lowe, Antibiotic-refractory Lyme arthritis Vaccine 27 (2009) 5740–5747. is associated with HLA-DR molecules that bind a Borrelia [95] H. Inaba, W. Martin, A.S. De Groot, S. Qin, L.J. De Groot, burgdorferi peptide, J. Exp. Med. 203 (2006) 961–971. Thyrotropin receptor epitopes and their relation to histocom- [114] J.A. McMurry, S.H. Gregory, L. Moise, D. Rivera, S. Buus, A.S. patibility leukocyte antigen-DR molecules in Graves' disease, De Groot, Diversity of Francisella tularensis Schu4 antigens J. Clin. Endocrinol. Metab. 91 (2006) 2286–2294. recognized by T after natural infections in [96] C.A.Weber,P.J.Mehta,M.Ardito,L.Moise,B.Martin,A.S. humans: identification of candidate epitopes for inclusion in De Groot, T cell epitope: friend or foe? Immunogenicity a rationally designed tularemia vaccine, Vaccine 25 (2007) of biologics in context, Adv. Drug Deliv. Rev. 61 (2009) 3179–3191. 965–976. [115] Y. Cheng, W.H. Prusoff, Relationship between the inhibition [97] L. Moise, C. Song, W.D. Martin, R. Tassone, A.S. De Groot, D.W. constant (K1) and the concentration of inhibitor which causes Scott, Effect of HLA DR epitope de-immunization of Factor VIII 50 per cent inhibition (I50) of an enzymatic reaction, in vitro and in vivo, Clin. Immunol. 142 (2012) 320–331. Biochem. Pharmacol. 22 (1973) 3099–3108. [98] M. Nielsen, O. Lund, NN-align. An artificial neural network- [116] S. Justesen, M. Harndahl, K. Lamberth, L.L. Nielsen, S. Buus, based alignment algorithm for MHC class II peptide binding Functional recombinant MHC class II molecules and high- prediction, BMC Bioinforma. 10 (2009) 296. throughput peptide-binding assays, Immunome Res. 5 (2009) 2. [99] H.G. Rammensee, Antigen presentation—recent develop- [117] R. Buchli, R.S. VanGundy, H.D. Hickman-Miller, C.F. Giberson, W. ments, Int. Arch. Allergy Immunol. 110 (1996) 299–307. Bardet,W.H.Hildebrand,Development and validation of a [100] D.R. Fooksman, G.K. Gronvall, Q. Tang, M. Edidin, Clustering fluorescence polarization-based competitive peptide-binding class I MHC modulates sensitivity of T cell recognition, J. assay for HLA-A*0201—a new tool for epitope discovery, Immunol. 176 (2006) 6673–6680. Biochemistry 44 (2005) 12491–12507. [101] A.J. Sant, F.A. Chaves, S.A. Jenks, K.A. Richards, P. Menges, [118] L.C. Wu, D.S. Tuot, D.S. Lyons, K.C. Garcia, M.M. Davis, J.M. Weaver, C.A. Lazarski, The relationship between Two-step binding mechanism for T-cell receptor recognition immunodominance, DM editing, and the kinetic stability of of peptide MHC, Nature 418 (2002) 552–556. MHC class II:peptide complexes, Immunol. Rev. 207 (2005) [119] S.T. Chang, D. Ghosh, D.E. Kirschner, J.J. Linderman, Peptide 261–278. length-based prediction of peptide–MHC class II binding, [102] C. Kuttler, A.K. Nussbaum, T.P. Dick, H.G. Rammensee, H. Bioinformatics 22 (2006) 2761–2767. Schild, K.P. Hadeler, An algorithm for the prediction of [120] Guidance for industry: immunogenicity assessment for therapeu- proteasomal cleavages, J. Mol. Biol. 298 (2000) 417–429. tic protein products, DHHS FDA Docket No. FDA-2013-D-0092, Td immunogenicity of biologics 553

Document No. 2013-03019, 2013, (https://federalregister.gov/ [138] C. Qu, V.A. Nguyen, M. Merad, G.J. Randolph, MHC class a/2013-03019). I/peptide transfer between dendritic cells overcomes poor [121] M.P. Baker, T.D. Jones, Identification and removal of cross-presentation by monocyte-derived APCs that engulf immunogenicity in therapeutic proteins, Curr. Opin. Drug dying cells, J. Immunol. 182 (2009) 3650–3659. Discov. Devel. 10 (2007) 219–227. [139] B.C. Schanen, A.S. De Groot, L. Moise, M. Ardito, E. McClaine, [122] L. Findlay, D. Eastwood, C. Ball, C.J. Robinson, C. Bird, M. W. Martin, V. Wittman, W.L. Warren, D.R. Drake III, Coupling Wadhwa, S.J. Thorpe, R. Thorpe, R. Stebbings, S. Poole, sensitive in vitro and in silico techniques to assess cross- Comparison of novel methods for predicting the risk of reactive CD4(+) T cells against the swine-origin H1N1 pro-inflammatory clinical infusion reactions during monoclonal influenza virus, Vaccine 29 (2011) 3299–3309. antibody therapy, J. Immunol. Methods 371 (2011) 134–142. [140] V. Dhir, M. Fort, A. Mahmood, R. Higbee, W. Warren, P. [123] G. Suntharalingam, M.R. Perry, S. Ward, S.J. Brett, A. Narayanan, V. Wittman, A predictive biomimetic model of Castello-Cortes, M.D. Brunner, N. Panoskaltsis, Cytokine cytokine release induced by TGN1412 and other therapeutic storm in a phase 1 trial of the anti-CD28 monoclonal antibody monoclonal antibodies, J. Immunotoxicol. 9 (1) (2012) 34–42. TGN1412, N. Engl. J. Med. 355 (2006) 1018–1028. [141] C. Giese, C.D. Demmler, R. Ammer, S. Hartmann, A. Lubitz, [124] C. Czerkinsky, G. Andersson, H.P. Ekre, L.A. Nilsson, L. Klareskog, L. Miller, R. Muller, U. Marx, A human lymph node in vitro— O. Ouchterlony, Reverse ELISpot assay for clonal analysis of challenges and progress, Artif. Organs 30 (2006) 803–808. cytokine production. I. Enumeration of gamma-interferon- [142] J.M. Moser, E.R. Sassano, C. Leistritz del, J.M. Eatrides, S. secreting cells, J. Immunol. Methods 110 (1988) 29–36. Phogat, W. Koff, D.R. Drake III, Optimization of a dendritic [125] C.C. Czerkinsky, L.A. Nilsson, H. Nygren, O. Ouchterlony, A. cell-based assay for the in vitro priming of naive human CD4+ Tarkowski, A solid-phase enzyme-linked immunospot (ELISpot) T cells, J. Immunol. Methods 353 (2010) 8–19. assay for enumeration of specific antibody-secreting cells, J. [143] R.A. Subbramanian, S. Basha, R.C. Brady, S. Hazenfeld, M.T. Immunol. Methods 65 (1983) 109–121. Shata, D.I. Bernstein, Age-related changes in magnitude and [126] R.T. Carson, D.A. Vignali, Simultaneous quantitation of 15 diversity of cross-reactive CD4+ T-cell responses to the novel cytokines using a multiplexed flowcytometricassay, J. Immunol. pandemic H1N1 influenza hemagglutinin, Hum. Immunol. 71 Methods 227 (1999) 41–52. (2010) 957–963. [127] B. Makitalo, M. Andersson, I. Arestrom, K. Karlen, F. Villinger, [144] H. Karasuyama, K. Mukai, Y. Tsujimura, K. Obata, Newly A. Ansari, S. Paulie, R. Thorstensson, N. Ahlborg, ELISpot discovered roles for basophils: a neglected minority gains new and ELISA analysis of spontaneous, mitogen-induced and respect, Nat. Rev. Immunol. 9 (2009) 9–13. antigen-specific cytokine production in cynomolgus and [145] M.M. Davis, J.D. Altman, E.W. Newell, Interrogating the rhesus macaques, J. Immunol. Methods 270 (2002) 85–97. repertoire: broadening the scope of peptide–MHC multimer [128] R. Basu, R.D. Hatton, C.T. Weaver, The Th17 family: analysis, Nat. Rev. Immunol. 11 (2011) 551–558. flexibility follows function, Immunol. Rev. 252 (2013) 89–103. [146] H.T. Maecker, J.P. McCoy, R. Nussenblatt, Standardizing [129] B. Foster, C. Prussin, F. Liu, J.K. Whitmire, J.L. Whitton, immunophenotyping for the Human Immunology Project, Nat. Detection of intracellular cytokines by flow cytometry, Rev. Immunol. 12 (2012) 191–200. Curr. Protoc. Immunol. John Wiley and Sons, 2007, (Chapter [147] S.Pan,T.Trejo,J.Hansen,M.Smart,C.S.David,HLA-DR4 6; Unit 6.24). (DRB1*0401) transgenic mice expressing an altered CD4-binding [130] P. Pala, T. Hussell, P.J. Openshaw, Flow cytometric mea- site: specificity and magnitude of DR4-restricted T cell response, surement of intracellular cytokines, J. Immunol. Methods 243 J. Immunol. 161 (1998) 2925–2929. (2000) 107–124. [148] Y.C. Kong, L.C. Lomo, R.W. Motte, A.A. Giraldo, J. Baisch, G. [131] L. Wooldridge, A. Lissina, D.K. Cole, H.A. van den Berg, D.A. Strauss, G.J. Hammerling, C.S. David, HLA-DRB1 polymorphism Price, A.K. Sewell, Tricks with tetramers: how to get the determines susceptibility to autoimmune thyroiditis in trans- most from multimeric peptide–MHC, Immunology 126 (2009) genic mice: definitive association with HLA-DRB1*0301 (DR3) 147–164. gene, J. Exp. Med. 184 (1996) 1167–1172. [132] A. Jaber, M. Baker, Assessment of the immunogenicity of [149] M. Shirai, T. Arichi, M. Nishioka, T. Nomura, K. Ikeda, K. different interferon beta-1a formulations using ex vivo T-cell Kawanishi, V.H. Engelhard, S.M. Feinstone, J.A. Berzofsky, assays, J. Pharm. Biomed. Anal. 43 (2007) 1256–1261. CTL responses of HLA-A2.1-transgenic mice specific for [133] M.M. Stickler, D.A. Estell, F.A. Harding, CD4+ T-cell epitope hepatitis C viral peptides predict epitopes for CTL of humans determination using unexposed human donor peripheral blood carrying HLA-A2.1, J. Immunol. 154 (1995) 2733–2742. mononuclear cells, J. Immunother. 23 (2000) 654–660. [150] A.X. Le, E.J. Bernhard, M.J. Holterman, S. Strub, P. Parham, [134] S. Delluc, G. Ravot, B. Maillere, Quantification of the E. Lacy, V.H. Engelhard, responses in preexisting CD4 T-cell repertoire specific for human erythro- HLA-A2.1 transgenic mice. Recognition of HLA alloantigens poietin reveals its immunogenicity potential, Blood 116 and utilization of HLA-A2.1 as a restriction element, J. (2010) 4542–4545. Immunol. 142 (1989) 1366–1371. [135] R.G. Higbee, A.M. Byers, V. Dhir, D. Drake, H.G. Fahlenkamp, [151] A. Didierlaurent, J.C. Ramirez, M. Gherardi, S.C. Zimmerli, M. J. Gangur, A. Kachurin, O. Kachurina, D. Leistritz, Y. Ma, R. Graf, H.A. Orbea, G. Pantaleo, R. Wagner, M. Esteban, J.P. Mehta, E. Mishkin, J. Moser, L. Mosquera, M. Nguyen, R. Parkhill, Kraehenbuhl, J.C. Sirard, Attenuated poxviruses expressing a S. Pawar, L. Poisson, G. Sanchez-Schmitz, B. Schanen, I. Singh, H. synthetic HIV protein stimulate HLA-A2-restricted cytotoxic Song,T.Tapia,W.Warren,V.Wittman, An immunologic model T-cell responses, Vaccine 22 (2004) 3395–3403. for rapid vaccine assessment — a clinical trial in a test tube, [152] S. Man, M.H. Newberg, V.L. Crotzer, C.J. Luckey, N.S. Altern. Lab. Anim. 37 (Suppl. 1) (2009) 19–27. Williams, Y. Chen, E.L. Huczko, J.P. Ridge, V.H. Engelhard, [136] C. Giese, A. Lubitz, C.D. Demmler, J. Reuschel, K. Bergner, U. Definition of a human T cell epitope from influenza A Marx, Immunological substance testing on human lymphatic non-structural protein 1 using HLA-A2.1 transgenic mice, Int. micro-organoids in vitro, J. Biotechnol. 148 (2010) 38–45. Immunol. 7 (1995) 597–605. [137] Y. Ma, L. Poisson, G. Sanchez-Schmitz, S. Pawar, C. Qu, G.J. [153] J. Charo, M. Sundback, A. Geluk, T. Ottenhoff, R. Kiessling, Randolph, W.L. Warren, E.M. Mishkin, R.G. Higbee, Assessing DNA immunization of HLA transgenic mice with a plasmid the immunopotency of Toll-like receptor agonists in an in expressing mycobacterial heat shock protein 65 results in HLA vitro tissue-engineered immunological model, Immunology class I- and II-restricted T cell responses that can be augmented 130 (2010) 374–387. by cytokines, Hum. Gene Ther. 12 (2001) 1797–1804. 554 V. Jawa et al.

[154] G.Y. Ishioka, J. Fikes, G. Hermanson, B. Livingston, C. Crimi, [169] T.Willinger,A.Rongvaux,T.Strowig,M.G.Manz,R.A.Flavell, M. Qin, M.F. del Guercio, C. Oseroff, C. Dahlberg, J. Improving human hemato-lymphoid-system mice by cytokine Alexander, R.W. Chesnut, A. Sette, Utilization of MHC class I knock-in gene replacement, Trends Immunol. 32 (2011) 321–327. transgenic mice for development of minigene DNA vaccines [170] M. Stickler, A.M. Valdes, W. Gebel, O.J. Razo, N. Faravashi, encoding multiple HLA-restricted CTL epitopes, J. Immunol. R. Chin, N. Rochanayon, F.A. Harding, The HLA-DR2 haplo- 162 (1999) 3915–3925. type is associated with an increased proliferative response [155] B.D. Livingston, M. Newman, C. Crimi, D. McKinney, R. to the immunodominant CD4(+) T-cell epitope in human Chesnut, A. Sette, Optimization of epitope processing interferon-beta, Genes Immun. 5 (2004) 1–7. enhances immunogenicity of multiepitope DNA vaccines, [171] A.S. Parker, W. Zheng, K.E. Griswold, C. Bailey-Kellogg, Vaccine 19 (2001) 4652–4660. Optimization algorithms for functional deimmunization of [156] T. Hanke, A.J. McMichael, M. Mwau, E.G. Wee, I. Ceberej, S. therapeutic proteins, BMC Bioinforma. 11 (2010) 180. Patel, J. Sutton, M. Tomlinson, R.V. Samuel, Development of [172] J.R. Cantor, T.H. Yoo, A. Dixit, B.L. Iverson, T.G. Forsthuber, a DNA-MVA/HIVA vaccine for Kenya, Vaccine 20 (2002) G. Georgiou, Therapeutic enzyme deimmunization by combi- 1995–1998. natorial T-cell epitope removal using neutral drift, Proc. Natl. [157] M.V. Ragni, F.A. Bontempo, J.H. Lewis, Disappearance of Acad. Sci. U. S. A. 108 (2011) 1272–1277. inhibitor to factor VIII in HIV-infected hemophiliacs with [173] R.G. Holgate, M.P. Baker, Circumventing immunogenicity in progression to AIDS or severe ARC, Transfusion 29 (1989) the development of therapeutic antibodies, IDrugs 12 (2009) 447–449. 233–237. [158] E.A. Clark, J.A. Ledbetter, How B and T cells talk to each [174] T.D. Jones, L.J. Crompton, F.J. Carr, M.P. Baker, Deimmunization other, Nature 367 (1994) 425–428. of monoclonal antibodies, Methods Mol. Biol. 525 (2009) [159] B.M. Reipert, M. Sasgary, R.U. Ahmad, W. Auer, P.L. Turecek, 405–423(xiv). H.P. Schwarz, Blockade of CD40/CD40 ligand interactions [175] A.S. De Groot, M. Goldberg, L. Moise, W. Martin, Evolutionary prevents induction of factor VIII inhibitors in hemophilic mice deimmunization: an ancillary mechanism for self-tolerance? but does not induce lasting immune tolerance, Thromb. Cell. Immunol. 244 (2006) 148–153. Haemost. 86 (2001) 1345–1352. [176] P.A. Warmerdam, S. Plaisance, K. Vanderlick,P.Vandervoort,K. [160] M.J. Holder, H. Wang, A.E. Milner, M. Casamayor, R. Brepoels, D. Collen, M. De Maeyer, Elimination of a human T-cell Armitage, M.K. Spriggs, W.C. Fanslow, I.C. MacLennan, C.D. region in staphylokinase by T-cell screening and computer Gregory, J. Gordon, et al., Suppression of apoptosis in normal modeling, Thromb. Haemost. 87 (2002) 666–673. and neoplastic human B lymphocytes by CD40 ligand is [177] A.S. De Groot, P.M. Knopp, W. Martin, De-immunization of independent of Bc1-2 induction, Eur. J. Immunol. 23 (1993) therapeutic proteins by T-cell epitope modification, Dev. 2368–2371. Biol. (Basel) 122 (2005) 171–194. [161] G.G. Klaus, M.S. Choi, E.W. Lam, C. Johnson-Leger, J. Cliff, [178] D.W. Scott, A.S. De Groot, Can we prevent immunogenicity of CD40: a pivotal receptor in the determination of life/death human protein drugs? Ann. Rheum. Dis. 69 (Suppl. 1) (2010) decisions in B lymphocytes, Int. Rev. Immunol. 15 (1997) i72–i76. 5–31. [179] K. Hellendoorn, T. Jones, J. Watkins, M. Baker, A. Hamilton, [162] J. Qian, L.C. Burkly, E.P. Smith, J.L. Ferrant, L.W. Hoyer, F. Carr, Limiting the risk of immunogenicity by identification D.W. Scott, C.C. Haudenschild, Role of CD154 in the and removal of T-cell epitopes (DeImmunisationTM), Cancer secondary immune response: the reduction of pre-existing Cell Int. 4 (2004) S20. splenic germinal centers and anti-factor VIII inhibitor titer, [180] V.P. Yeung, J. Chang, J. Miller, C. Barnett, M. Stickler, F.A. Eur. J. Immunol. 30 (2000) 2548–2554. Harding, Elimination of an immunodominant CD4+ T cell [163] L. Bi, A.M. Lawler, S.E. Antonarakis, K.A. High, J.D. Gearhart, epitope in human IFN-beta does not result in an in vivo H.H. Kazazian Jr., Targeted disruption of the mouse factor response directed at the subdominant epitope, J. Immunol. VIII gene produces a model of haemophilia A, Nat. Genet. 10 172 (2004) 6658–6665. (1995) 119–121. [181] S.P. Cobbold, S.X. Qin, H. Waldmann, Reprogramming the [164] A. Mangalam, M. Rodriguez, C. David, A new humanized HLA immune system for tolerance with monoclonal antibodies, transgenic mouse model of multiple sclerosis expressing class Semin. Immunol. 2 (1990) 377–387. II on mouse CD4 T cells, Ann. N. Y. Acad. Sci. 1103 (2007) [182] D. Winsor-Hines, C. Merrill, M. O'Mahony, P.E. Rao, S.P. 112–117. Cobbold, H. Waldmann, D.J. Ringler, P.D. Ponath, Induction [165] M.A. Brehm, A. Cuthbert, C. Yang, D.M. Miller, P. DiIorio, J. of immunological tolerance/hyporesponsiveness in baboons Laning, L. Burzenski, B. Gott, O. Foreman, A. Kavirayani, M. with a nondepleting CD4 antibody, J. Immunol. 173 (2004) Herlihy, A.A. Rossini, L.D. Shultz, D.L. Greiner, Parameters 4715–4723. for establishing humanized mouse models to study human [183] S.C. Jordan, M. Toyoda, A.A. Vo, Regulation of immunity and immunity: analysis of human hematopoietic stem cell en- inflammation by intravenous immunoglobulin: relevance to graftment in three immunodeficient strains of mice bearing solid organ transplantation, Expert. Rev. Clin. Immunol. 7 the IL2rgamma(null) mutation, Clin. Immunol. 135 (2010) (2011) 341–348. 84–98. [184] G. Zandman-Goddard, A. Krauthammer, Y. Levy, P. Langevitz, Y. [166] F.Ishikawa,Y.Saito,S.Yoshida,M.Harada,L.D.Shultz,The Shoenfeld, Long-term therapy with intravenous immunoglobulin differentiative and regenerative properties of human hema- is beneficial in patients with autoimmune diseases, Clin. Rev. topoietic stem/progenitor cells in NOD-SCID/IL2rgamma Allergy Immunol. 42 (2) (2012) 247–255. (null) mice, Curr. Top. Microbiol. Immunol. 324 (2008) [185] A.C. Dykes, I.D. Walker, G.D. Lowe, R.C. Tait, Combined 87–94. prednisolone and intravenous immunoglobulin treatment for [167] A.K. Wege, M.W. Melkus, P.W. Denton, J.D. Estes, J.V. acquired factor VIII inhibitors: a 2-year review, Haemophilia 7 Garcia, Functional and phenotypic characterization of the (2001) 160–163. humanized BLT mouse model, Curr. Top. Microbiol. Immunol. [186] D. Provan, Characteristics of immune thrombocytopenic 324 (2008) 149–165. purpura: a guide for clinical practice, Eur. J. Haematol. [168] P. Reddy, R. Negrin, G.R. Hill, Mouse models of bone marrow Suppl. (2009) 8–12. transplantation, Biol. Blood Marrow Transplant. 14 (2008) [187] D.B. Cines, H. Liebman, R. Stasi, Pathobiology of secondary 129–135. immune thrombocytopenia, Semin. Hematol. 46 (2009) S2–S14. Td immunogenicity of biologics 555

[188] P. Lozeron, D. Adams, Advances in the treatment of chronic [195] S. van der Marel, A. Majowicz, K. Kwikkers, R. van inflammatory demyelinating neuropathies in 2010, J. Neurol. Logtenstein, A.A. te Velde, A.S. De Groot, S.L. Meijer, S.J. 258 (2011) 1737–1741. van Deventer, H. Petry, D.W. Hommes, V. Ferreira, Adeno- [189] M. Mahdi-Rogers, Y.A. Rajabally, Overview of the pathogen- associated virus mediated delivery of Tregitope 167 amelio- esis and treatment of chronic inflammatory demyelinating rates experimental colitis, World J. Gastroenterol. 18 (2012) polyneuropathy with intravenous immunoglobulins, Biologics 4288–4299. 4 (2010) 45–49. [196] D.T. Mytych, S. La, T. Barger, J. Ferbas, S.J. Swanson, [190] N.J. Mendelsohn, Y.H. Messinger, A.S. Rosenberg, P.S. The development and validation of a sensitive, dual-flow Kishnani, Elimination of antibodies to recombinant enzyme cell, SPR-based biosensor immunoassay for the detection, in Pompe's disease, N. Engl. J. Med. 360 (2009) 194–195. semi-quantitation, and characterization of antibodies to [191] R.D. Garman, K. Munroe, S.M. Richards, Methotrexate darbepoetin alfa and epoetin alfa in human serum, J. reduces antibody responses to recombinant human alpha- Pharm. Biomed. Anal. 49 (2009) 415–426. galactosidase A therapy in a mouse model of Fabry disease, [197] M. Wadhwa, R. Thorpe, Unwanted immunogenicity: les- Clin. Exp. Immunol. 137 (2004) 496–502. sons learned and future challenges, Bioanalysis 2 (2010) [192] Y. Su, R. Rossi, A.S. De Groot, D.W. Scott, Regulatory T cell 1073–1084. epitopes (Tregitopes) in IgG induce tolerance in vivo and lack [198] J.M. Cruzado, O. Bestard, J.M. Grinyo, New immunosuppres- immunogenicity per se, J. Leukoc. Biol. 94 (2) (2013) 377–383. sive protocols with the advent of novel biological drugs, [193] L.P. Cousens, Y. Su, E. McClaine, X. Li, F. Terry, R. Smith, J. Transplantation 88 (2009) S20–S23. Lee, W. Martin, D.W. Scott, A.S. De Groot, Application [199] A.S. De Groot, R. Tassone, L.P. Cousens, F. Terry, R. Martin, of IgG-derived natural Treg epitopes (IgG tregitopes) to M. Ardito, W. Martin, In silico prediction of HLA-DP and ?DQ antigen-specific tolerance induction in a murine model of epitope content is poorly correlated with clinical immunoge- type 1 diabetes, J. Diabetes Res. 2013 (2013) 621693. nicity of therapeutic proteins, manuscript in preparation [194] F. Mingozzi, K.A. High, Immune responses to AAV vectors: (2013), Abstract presented at the American Association of overcoming barriers to successful gene therapy, Blood 122 (1) Pharmaceutical Scientists (AAPS) National Biotechnology (2013) 23–26. Conference (NBC), San Diego, CA, USA. May 20-22, 2013.